|Home | About | Journals | Submit | Contact Us | Français|
Antiviral therapy reduces symptom duration and hospitalization risk among previously healthy and chronically ill children infected with seasonal influenza. The impact of oseltamivir on outcomes of hospitalized children is unknown. The primary objective of this study was to determine whether oseltamivir improves outcomes of critically ill children hospitalized with influenza.
We performed a retrospective cohort study of children admitted to a pediatric intensive care unit with influenza during 6 consecutive winter seasons (2001-2007). We used the Pediatric Health Information System (PHIS) database, which contains resource utilization data from 41 children's hospitals. We matched oseltamivir-treated and oseltamivir-non-treated patients by the probability of oseltamivir exposure using a propensity score we derived from patient and hospital characteristics. We subsequently compared the outcomes of critically ill children treated with oseltamivir within 24 hours of admission with propensity score matched children who were not treated with oseltamivir.
We identified 1,257 children with influenza infection, 264 of whom were treated with oseltamivir within 24 hours of hospital admission. Multivariable analysis of 252 oseltamivir-treated patients and 252 propensity score- matched untreated patients demonstrated that patients treated with oseltamivir experienced a 18% reduction in total hospital days (Time Ratio: 0.82, p=0.02) whereas intensive care unit stay, in-hospital mortality and readmission rates did not differ.
For critically ill children infected with seasonal influenza, treatment with oseltamivir within 24 hours of hospitalization was associated with a shorter duration of hospital stay. Additional study is needed to determine the impact of delayed initiation of oseltamivir on clinical outcomes.
Each year in the United States, one in every 100 children less than 5 years of age is hospitalized with influenza or influenza-related complications.1 Over the past 5 years, researchers have shown the potential severity of influenza. In 2004, pediatric influenza-related deaths became nationally reportable; since that time 43-88 deaths have been reported each year.2, 3 Although children with chronic conditions are at increased risk of influenza related complications4, approximately half of fatal pediatric infections occur in previously healthy children.5-7
In 2009, the emergence and rapid spread of a novel pandemic influenza strain underscored the urgent need for a better understanding of the effectiveness of antiviral medications in the treatment of patients infected with influenza. More than 10 clinical trials of anti-influenza medications have been conducted in non-hospitalized patients. The majority of these trials demonstrated that antiviral medications, including oseltamivir and amantadine, shortened the duration of fever and influenza symptoms when initiated within 48 hours of symptom onset.8-15 In addition, antiviral therapy reduced the rate of influenza-related respiratory complications and hospitalizations in both previously healthy and chronically ill ambulatory patients.8, 11, 13, 16-18 Antiviral therapy has also been shown to reduce mortality among hospitalized adults. 19,20 Although influenza is a common cause of pediatric hospitalization, little is known about the impact of antiviral medications on the course of illness among children with influenza. We undertook this study to examine the effectiveness of oseltamivir to alter outcomes of critically ill children hospitalized with influenza.
We performed a retrospective cohort study of children admitted to an intensive care unit for treatment of influenza at one of 41 children's hospitals in the United States. We compared the outcomes of critically ill children treated with oseltamivir to propensity score matched children who did not receive oseltamivir.
Data for this study were obtained from the Pediatric Health Information System (PHIS). PHIS is a national administrative database containing resource utilization data from 41 freestanding, tertiary care children's hospitals. PHIS-participating hospitals account for 20% of all tertiary care general (rather than subspecialty) children's hospitals. They are located in 23 different states plus the District of Columbia; no more than one participating hospital is present in a specific metropolitan area. These hospitals are affiliated with the Child Health Corporation of America (Shawnee Mission, KS), a business alliance of children's hospitals. Data quality and reliability are assured through a joint effort between the Child Health Corporation of America and participating hospitals. For the purposes of external benchmarking, participating hospitals provide discharge data including patient demographics, diagnoses, and procedures. Billing data are also available on a daily basis that details all of the drugs, radiologic imaging studies, laboratory tests, and supplies charged to each patient. Daily room charges allow for determination of patient location (e.g. neonatal intensive care unit, pediatric intensive care unit, medical / surgical ward). Systematic monitoring, including bimonthly coding consensus meetings, coding consistency reviews, and quarterly data quality reports, occur on an ongoing basis to ensure data quality. Analyses of PHIS data have been published in many peer-reviewed journals studying research topics spanning a wide variety of pediatric and pediatric sub-subspecialty disciplines21-26.
The study sample consisted of patients age 0 to 21 years who (1) had an International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) discharge diagnosis code of influenza (487.0, 487.1, or 487.8 in any position on discharge); (2) were discharged from any of the 41 participating hospitals between October and April during the years 2001 to 2007 (limited to patients discharged from hospitals with billing data during the time period); (3) had a charge for an influenza test within the first 48 hours of admission (test results are not currently available in PHIS); and (4) were directly admitted to an intensive care unit from home or transferred from another institution. We excluded any patient who received an influenza antiviral medication other than oseltamivir (i.e. zanamivir, amantadine, or rimantadine; N=989) or who had been hospitalized in the preceding 14 days (N=295). Only the first influenza encounter was included for all patients who had multiple influenza-related admissions during the same season. Patients hospitalized at PHIS hospitals with significant data issues (e.g. inaccurate day of service) identified by PHIS during regularly scheduled data quality checks were also excluded (N=236).
The primary exposure of interest was treatment with oseltamivir within the first 48 hours of hospital admission, as defined by a charge for oseltamivir on the calendar day of hospital admission (hospital day 0) or the first 24-hour calendar day of hospitalization (hospital day 1). Untreated patients were defined by the absence of a charge for oseltamivir at any time during the hospitalization or a charge for oseltamivir on hospital day 2 or greater.
The primary outcome was length of hospital stay (measured in days). Secondary outcomes included (1) length of intensive care unit stay (2) in-hospital mortality, and (3) readmission within 7 days from hospital discharge.
We collected information about patient age, gender, race, influenza season, and census region. The presence of a co-morbid condition was assessed using an ICD-9-CM based diagnostic classification system for pediatric complex chronic conditions.27 The categories include neuromuscular, cardiovascular, respiratory, renal, gastrointestinal, hematologic or immunologic, malignancy, and other congenital defect conditions. Information regarding influenza vaccination status is not included in this database and could not be assessed. We also collected information on supportive therapies on hospital day 0 (defined as at least one hospital charge in the billing record). Supportive therapies included mechanical ventilation, high-frequency ventilation, other assisted ventilation (including CPAP, Bi-Pap, and any non-invasive ventilation support), use of vasoactive medications, nitric oxide, and supplemental oxygen.
To account for potential confounding by observed baseline covariates, we matched treated patients with untreated patients using a propensity score we derived to estimate the likelihood of receiving oseltamivir (i.e. clinically eligible to receive oseltamivir) based on the presence or absence of baseline covariates. This approach can balance covariates between the treated and untreated groups better than other strategies such as conventional multivariable methods.28,29, 30 Propensity scores were calculated from a multivariable logistic regression that modeled receipt of the drug by the following covariates: age, sex, race, season, supportive therapies at admission, and the presence of individual complex chronic conditions. Two additional variables were included in the propensity score model to adjust for the severity of illness. First, we included the PHIS expected mortality variable (risk of mortality), a risk adjustment measure based upon the risk of mortality derived from Thomson Reuter's national database and 3M's All-Patient-Refined Diagnosis-Related Group (APR-DRG) classification system. Second, we derived a variable to indirectly assess the intensity of care delivered during the first 24-hour hospital calendar day (hospital day 1). Hospital bills from hospital day 1 were reviewed and all unique charges (i.e. medications, radiology, laboratory tests) were counted for each patient, with the assumption that the severity of illness would be directly associated with the number of unique charges.
We matched oseltamivir treated patients to untreated patients within each hospital. We chose to match by hospital to account for the between hospital variability in the utilization rate of oseltamivir.31 Each treated patient was matched to one untreated control patient using nearest-neighbor matching with a caliper set at one quarter of the standard deviation of the logit of the propensity scores.32 In order to make within hospital contrasts between the matched sets, only hospitals with a >=10% oseltamivir utilization rate of were included in the analysis. To determine whether the propensity score was successful in balancing the covariates between cases and matched controls, we performed pair-wise comparisons. We used accelerated failure time models to compare intensive care unit length of stay and total hospital length of stay between treated and untreated patients.33 Death was included in the model as a covariate to distinguish between two groups with a priori different length of stays. To account for unobserved heterogeneity among patients, we added a frailty term with gamma distribution to the model. For binomial outcomes, we used logistic link models to determine the difference in outcomes while adjusting for covariates. Covariates that were determined to be well balanced were not included in the models. In order to further explore the effect of death on our primary outcome, length of stay, we excluded all patients who died from the dataset, re-matched treated and untreated patients, and repeated the analysis. All statistical analyses were performed using the statistical software SAS 9.1 (SAS Institute, Inc., Cary, NC) and Stata 10.0 (Stata Corp., College Station, TX), and p<0.05 was considered statistically significant.
The protocol for the conduct of this study was reviewed and approved by The Children's Hospital of Philadelphia Committee for the Protection of Human Subjects with a waiver of informed consent.
We identified a total of 17,072 children with influenza during the study period (Figure 1). After exclusions (such as treatment with influenza antiviral medications other than oseltamivir, hospitalization in the preceding 14 days, and admission to a non-intensive care unit hospital ward) a total of 1,257 children were included in our study cohort. The median age of patients was 1.7 years (interquartile range, 0.45 to 5.6) and 39% had one or more complex chronic conditions.
Overall, 264 of 1,257 patients (21%) were treated with 1 or more doses of oseltamivir within the first 24 hour calendar day of hospital admission. The age distribution of treated and untreated patients differed; younger patients (<2 years) were less likely to have received oseltamivir (p<0.001) (Table 1). Oseltamivir utilization also varied by year; 2 percent of treated patients received the drug in 2001-2002 whereas 36 percent of treated patients received the drug in 2006-2007 (p<0.001). There were fewer treated patients located in the Western region of the United States (9%) as compared with the Northeast (22%), South (35%), and North Central U.S. (34%), (p<0.001; data not shown).
Patients treated with oseltamivir were more likely to have a neuromuscular condition (25% vs. 16%, p<0.001), metabolic condition (5% vs. 2%, p=0.038), or other congenital or genetic defect (9% vs. 6%, p=0.049) than untreated patients and were more likely to require mechanical ventilation (11% vs. 5%, p<0.001) or vasoactive medications (23% vs. 13%, p<0.001) on hospital day 0. As compared with untreated patients, oseltamivir treated patients had a greater predicted risk of mortality (0.43% vs. 0.02%) and a greater total number of unique charges (28.5 vs. 20) on hospital day 1 (p<0.001).
In the propensity score matched analysis, 95% of treated patients were matched to appropriate untreated patients. One patient from 1 hospital was excluded due to a low oseltamivir utilization rate (<10%), while eleven treated patients from 6 hospitals failed to match to appropriate control patients and were excluded. After successfully matching 95% of the treated patients, there was one statistically significant difference between matched treated and untreated patients (Table 1) in contrast to the unmatched analysis described above. Patient age was significantly different between the treated and untreated patients and was included in the final model. In addition, the variable neuromuscular co-morbid condition was considered for inclusion in the model. In the propensity matched analysis, we found that treatment with oseltamivir was associated with a shorter length of hospital stay (Table 2). Length of stay (in days) for treated patients was 18% shorter than propensity score matched untreated patients while controlling for death (Time Ratio: 0.82, 95% Confidence Interval. 0.69, 0.97, p=0.02). The addition of death to the model had no impact on the time ratio or p-value. We excluded patients who died from this analysis; our finding persisted with an 18% shorter length of stay for treated patients (Time Ratio: 0.82, 95% Confidence Interval. 0.71, 0.95, p=0.007). There was no difference between treated and untreated patients in length of ICU stay (p=0.51), the in-hospital mortality rate (p=0.67), or readmission rate within 7 days from discharge (p=0.42).
We found that critically ill children treated with oseltamivir had a shorter duration of total hospital stay when compared with matched patients who did not receive antiviral treatment within 24 hours of hospital admission. The duration of hospital stay was approximately 18% shorter for critically ill children treated with oseltamivir within 24 hours of hospitalization as compared with those whom were either untreated or received oseltamivir more than 24 hours after hospital admission. This is the first study to demonstrate a benefit of oseltamivir in critically ill children hospitalized with influenza.
Prior studies have demonstrated improved outcomes for non-hospitalized patients who received antiviral therapy. Oseltamivir treatment has been associated with reduced duration of illness, rates of influenza-related complications, and subsequent hospitalization among previously healthy, influenza-infected children8 and adults.17, 34 Using health insurance claims data, Piedra and colleagues35 recently demonstrated that oseltamivir therapy improved the outcomes of influenza-infected children with chronic medical conditions. Chronically ill outpatients who received oseltamivir within 1 day of influenza diagnosis had lower rates of respiratory complications, otitis media, and all-cause hospitalizations as compared to untreated children. Recently published data suggest initiation of antiviral therapy within 48 hrs of symptom onset, as compared with either delayed or no antiviral therapy, was associated with better clinical outcomes among adults hospitalized 2009 H1N1 influenza.18
Although children treated with oseltamivir had a shorter total length of hospital stay, we did not find a difference in the secondary outcomes examined such as duration of ICU stay, mortality, or readmission within 7 days. We did not power the study detect a difference in mortality, as the mortality rate among children hospitalized with influenza has historically been relatively low.5 Because the median duration of ICU stay was 4 days for both treated and untreated patients and ICU stay was measured in days, we suspect we were also underpowered to detect modest differences in this outcome. Because oseltamivir prevents viral replication we hypothesize that the drug may have minimal effect in altering the course of acute inflammatory reaction associated with severe influenza. Oseltamivir may hasten the time to resolution of other symptoms and thereby shortening the duration of non-ICU hospital stay. It is unclear why there was no difference between treated and untreated patients in the readmission rate.
Our study had several limitations. First, the use of administrative data might have led to misclassification of the principal exposure of interest (oseltamivir treatment) and important covariates (such as severity of illness measures), however we believe this bias would be non-differential. We recognize that we might not have fully adjusted for all differences in severity of illness. However, any residual confounding by indication would have biased our findings toward oseltamivir having less effect on the outcome of interest; thus, we believe that the difference in length of stay would be even greater in favor of the oseltamivir treated group. Selection bias (e.g. the use of ICD-9 codes for identifying influenza infected patients) might have led to under-ascertainment of influenza-infected patients; however this bias would have also led to a reduction in the difference in the duration of hospital stay among treated and untreated patients. Finally, we were unable to determine whether the time interval between symptom onset and initiation of oseltamivir influenced the impact of antiviral therapy on patient outcomes. Prior reports have suggested that antiviral medications may have negligible effect if begun greater than 48 hours after symptom onset.36 However, it is unlikely that all patients had symptoms for only 24 hours and therefore the effect size is likely to have been greater had treatment been initiated prior to hospital admission.
In summary, oseltamivir treatment reduced the duration of hospital stay among critically ill children with influenza infection when begun within 24 hours of hospital admission. This finding can assist clinical decision making to improve patient outcomes. In the event of critical shortages, our results suggest that severely ill children should be given priority to receive antiviral treatment. Future studies are needed to assess the effectiveness of antiviral medications upon specific populations of critically ill patients and less seriously ill hospitalized children.
This work was supported by a grant from the National Institute of Child Health and Human Development (5R03HD55966-2)
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.