Pediatric hospitalizations for pneumonia have declined since the introduction of routine childhood pneumococcal vaccination.1 Hospitalizations for complicated pneumonia (pneumonia with parapneumonic effusion or empyema), however, are increasing.2,3 The increasing incidence of complicated pneumonia is important as affected children often require invasive procedures and prolonged hospitalization.4- 10
Another factor that may contribute to disease severity in both complicated and uncomplicated pneumonia is the presence of coinfection with bacteria and viruses. Respiratory viruses predispose individuals to the development of bacterial pneumonia,11- 13 and in vitro and animal studies suggest that the individual host effects of both the viral and bacterial pathogens may act in concert to enhance the virulence of both pathogens in coinfected individuals.14 In the case of influenza, bacterial coinfection has been recognized as a complicating factor for nearly a century, and it is now thought that the majority of the more than 500 000 deaths in the United States attributed to the 1918 influenza pandemic resulted from bacterial coinfection.15 More recently, surveillance studies of pediatric influenza-related hospitalizations in the United States and Canada demonstrate that confirmed or suspected bacterial coinfection occurred in up to 23% of hospitalizations for influenza,16- 18 whereas mortality studies implicated bacterial coinfection in 24% to 40% of influenza-associated pediatric deaths.19- 21 Moreover, coinfected children required more frequent intensive care unit admissions as compared with those with influenza or bacterial infection alone.17,18 Thus, the presence of coinfection with bacteria and influenza appears to be strongly associated with increased disease severity.
Finally, although the association between complicated bacterial pneumonia and influenza has been recognized,22,23 no studies to our knowledge have specifically examined the impact of influenza coinfection on disease severity for children with complicated bacterial pneumonia, a group potentially at extremely high risk for poor outcomes. Thus, this study sought to address this issue through the use of a large retrospective cohort of children with complicated pneumonia from 40 children's hospitals nationwide. The primary objective was to determine whether influenza coinfection is independently associated with worse clinical outcomes.
Data for this multicenter retrospective cohort study were obtained from the Pediatric Health Information System, which contains resource utilization data from 40 freestanding children's hospitals. Participating hospitals are located in noncompeting markets of 27 states plus the District of Columbia and account for 20% of all tertiary care children's hospitals. These hospitals provide discharge data including patient demographic characteristics, diagnoses, and procedures. Billing data detail all drugs, radiologic imaging studies, laboratory tests, and supplies charged to each patient. Data are deidentified prior to inclusion in the database; however, encrypted medical record numbers allow for tracking individual patients across admissions. The Child Health Corporation of America, Shawnee Mission, Kansas, and participating hospitals jointly ensure data quality as described previously.24,25 In accordance with the Common Rule26 and the policies of the Children's Hospital of Philadelphia Institutional Review Board, this research, using a deidentified data set, was not considered human subjects research.
Children aged 18 years and younger requiring a pleural drainage procedure for complicated pneumonia were eligible if they were discharged from any of the 40 participating hospitals between January 1, 2004, and June 30, 2009. Study participants met the following criteria: (1) discharge diagnosis of pneumonia (International Classification of Diseases, Ninth Revision27 [ICD-9 ] discharge diagnosis codes 480.x-483.x, 485.x-487.x); (2) discharge diagnosis of pleural effusion (ICD-9 codes 510.0, 510.9, 511.0, 511.1, or 511.9); and (3) billing charge for antibiotics within 2 days of hospitalization. Additionally, the primary discharge diagnosis had to be either pneumonia or pleural effusion. Patients were excluded if they did not undergo a pleural fluid drainage procedure or if they had a chronic comorbid condition that would predispose them to severe, recurrent, or health care–associated pneumonia. These chronic comorbid conditions included underlying malignant, cardiac, or neurological diseases and were identified using a previously reported classification scheme.28
Patients were classified as having complicated pneumonia caused by Staphylococcus aureus, Streptococcus pneumoniae, Streptococcus pyogenes, or other bacteria/not specified. Patients with S aureus were identified with the ICD-9 codes V09.8, V12.04, 041.11-2, 038.11-2, and 482.41-2. Methicillin-resistant S aureus was further identified by the following codes: V09.8, V12.04, 041.12, 038.12, or 482.42. Patients with S pneumoniae were identified with the ICD-9 codes V03.82, 038.2, 481, or V06.6. Streptococcus pyogenes was identified with the ICD-9 codes 482.31 or 041.01. Pleural drainage procedures were identified using ICD-9 procedure codes for thoracentesis (34.91), chest tube placement (34.04), video-assisted thoracoscopic surgery (34.21), and thoracotomy (34.02 or 34.09). Fibrinolysis was defined as receipt of urokinase, streptokinase, or alteplase within 2 days of initial chest tube placement.
Billing data were used to classify receipt of mechanical ventilation (invasive and noninvasive methods, including continuous positive airway pressure) and medications, including antibiotics, on the first day of hospitalization. Antiviral therapy for influenza included oseltamivir phosphate, zanamivir, amantadine hydrochloride, and rimantadine hydrochloride. Blood product transfusions included administration of packed red blood cells, cryoprecipitate, fresh frozen plasma, or platelets. Vasoactive infusions included dobutamine hydrochloride, dopamine, epinephrine, norepinephrine bitartrate, vasopressin, and milrinone.
Measured outcomes included intensive care unit admission, receipt of mechanical ventilation, receipt of vasoactive infusions, receipt of blood product transfusions, in-hospital death, readmission within 14 days of hospital discharge, total hospital length of stay, and total cost of hospitalization.
The primary exposure of interest was influenza coinfection, identified by ICD-9 codes (487, 487.0, 487.1, 487.8, 488, or V04.81). Additional model covariates included age, race, sex, empirical antibiotic therapy, initial pleural fluid drainage procedure, causative bacterium, and hospital.
Variables were summarized using frequencies and percentages for categorical variables and using median, interquartile range, and range for continuous variables. Outcomes by influenza coinfection were compared using χ2 tests for categorical variables and Wilcoxon rank sum tests for continuous variables.
Multivariable analysis was performed to account for potential confounding by observed baseline variables. For dichotomous outcome variables, modeling consisted of logistic regression using the method of generalized estimating equations to account for hospital clustering. For continuous variables, a mixed model approach was used, treating hospital as a random effect. Since length of stay and cost were right-skewed, log transformation was applied to these outcome variables prior to modeling. Odds ratios and 95% confidence intervals were reported for comparison of dichotomous outcomes and the adjusted means and 95% confidence intervals were reported for continuous outcomes after appropriate back transformation. All analyses were clustered by hospital. Analyses were performed using SAS version 9.2 statistical software (SAS Institute, Inc, Cary, North Carolina). A 2-tailed P < .05 was considered statistically significant.
To account for potential confounding by indication for influenza testing, analyses were repeated while limiting the cohort to patients with a billing charge for influenza testing. Finally, unadjusted outcomes of patients with and without influenza were compared after stratifying by causative bacteria to identify important subgroup differences.
During the study period, 9680 patients had complicated pneumonia. Subjects were excluded if they did not have a pleural fluid drainage procedure (n = 5798) or if they had an underlying chronic comorbid condition (n = 500). The remaining 3382 children with complicated pneumonia were included; 105 (3.1%) of these children had influenza coinfection. The median patient age was 4.1 years (interquartile range, 2.1-6.9 years). Forty children (1.2%) had a pneumonia-associated emergency department visit or hospital discharge in the 7 days before hospitalization for complicated pneumonia, although no patient with influenza had a recent emergency department visit; the difference was not statistically significant compared with those without influenza (P = .64). Most patients received empirical therapy with a cephalosporin in combination with either vancomycin hydrochloride or clindamycin hydrochloride. The primary pleural drainage strategy was chest tube placement without fibrinolysis; most pleural fluid drainage procedures were performed within 2 days of hospitalization. Bacterial causes were specified in approximately one-third of cases (Table 1); commonly identified pathogens included S pneumoniae and S aureus. Methicillin-resistant S aureus accounted for most S aureus isolates (n = 282; 72.1%); differences in the proportion of methicillin-resistant isolates among all S aureus isolates between patients with influenza (66.7%) and without influenza (72.5%) did not reach statistical significance (P = .54).
Characteristics of patients with and without influenza coinfection are compared in Table 1. No significant differences in age, race, sex, and primary payer existed between the 2 groups. Staphylococcus aureus was the most commonly identified bacterial pathogen in children with influenza coinfection, while S pneumoniae was identified most commonly among children without influenza coinfection (Table 1). Among patients without influenza coinfection, S pneumoniae was the causative bacteria in 656 of 1159 patients (56.6%) with a specified bacterial cause. In contrast, S aureus was the causative bacteria in 24 of 42 patients (57.1%) with influenza and a specified bacterial cause. Only 20 children with influenza coinfection (19.0%) and 7 without coinfection (0.2%) received antiviral agents.
In an unadjusted analysis, patients with influenza coinfection had a significantly higher use of the medical interventions, including intensive care unit admission, mechanical ventilation, vasoactive infusions, and blood product transfusions, as well as a longer hospital stay and higher mortality and total hospital costs (Table 2). In multivariable analysis, influenza coinfection remained associated with higher odds of intensive care unit admission and receipt of mechanical ventilation, vasoactive infusions, and blood product transfusions, higher costs, and a longer hospital stay (Table 3). Children with influenza coinfection were less likely to require readmission, although there was a trend toward significantly higher odds of mortality for patients with coinfection (Table 3).
Influenza testing was performed in 551 patients (16.8%) without influenza coinfection. In a subanalysis, the outcomes of those with influenza coinfection were consistently worse than those tested for influenza but without documented coinfection. However, statistically significant differences occurred only in the requirement for mechanical ventilation (adjusted odds ratio = 1.11; 95% confidence interval, 1.10-1.12) and total hospital cost (adjusted difference = $5099; 95% confidence interval, $2047-$8436) in the coinfected group compared with those without coinfection. In the subanalysis stratified by bacteria, unadjusted outcomes were generally worse for patients with influenza in the subgroups of patients with S aureus and with no specified bacteria (Table 4). Relatively few children in the S pneumoniae subgroup had influenza coinfection (n = 15), and although several of the outcomes appeared worse for the coinfected children, none reached statistical significance.
In this national, multicenter study that included more than 3000 children with complicated pneumonia, documented influenza coinfection occurred in 3.1% of children and was associated with worse clinical outcomes. While S pneumoniae was the most commonly identified bacterial organism overall, S aureus was the most commonly identified bacterium among children with influenza coinfection.
Several studies have demonstrated an association between influenza and complicated pneumonia22,23; however, this is the first study to our knowledge to specifically examine the impact of influenza coinfection on outcomes for children with complicated pneumonia. Although influenza infection occurred uncommonly in this cohort, children coinfected with influenza had an increased likelihood of requiring intensive care unit admission and supportive care measures, had a longer hospital stay, and had higher total hospital costs as compared with children without influenza coinfection. There was also a trend toward higher mortality among coinfected children. Several recent studies have documented similar poor outcomes, including death, among children with primarily uncomplicated pneumonia and influenza coinfection.15,17,20 Comparable to the findings in our study, this observation also supports previous work suggesting that influenza and bacterial pathogens, when present in the same individual, result in infections more severe than that expected for either pathogen alone.14 In contrast, children without coinfection had a greater likelihood of readmission. This difference may be attributed to longer hospitalizations among coinfected children, allowing for more complete recovery before discharge.
Despite the well-established link between influenza and both S pneumoniae and S aureus infections, S aureus occurred most commonly among coinfected individuals in our study, disproportionately accounting for nearly two-thirds of all coinfections in which a bacterial pathogen was identified. Hospitalizations for invasive staphylococcal infections, including pneumonia, have increased in the United States—a finding accounted for almost exclusively by methicillin-resistant S aureus,29 no doubt as a result of the increasing prevalence of community-associated methicillin-resistant S aureus.30 This pathogen, which accounts for more than 70% of staphylococcal infections in some regions,31- 33 has been linked to severe pneumonia, including complications such as empyema, abscess formation, and necrosis.34,35 Recent studies of complicated pneumonia (with or without coinfection) have documented a similarly high prevalence of methicillin resistance (55%-100%) among culture-proven S aureus isolates.17,18,20,21,34,35 Moreover, recent reports of influenza-related mortality increasingly implicate S aureus as one of the most commonly identified bacterial pathogens leading to death among children with influenza,20 including during the 2009 novel influenza A(H1N1) pandemic.36 These findings along with the observations of our study suggest that S aureus may contribute disproportionately to complicated pneumonia-associated morbidity, a factor that may become even more important as uptake of newer pneumococcal conjugate vaccine formulations becomes widespread.
Because S aureus was more commonly identified among coinfected children, the increased severity observed for coinfection in this study could be attributed to the increased prevalence of S aureus. To address this concern, we conducted a subanalysis comparing outcomes for children with and without influenza coinfection while stratifying by bacterial pathogen. Children with influenza coinfection either with S aureus or without a specified bacterium had significantly worse clinical outcomes than children without coinfection. Thus, although staphylococcal pneumonia may contribute to overall increased pneumonia severity, it does not, in and of itself, explain the poor clinical outcomes observed for coinfected children. Our findings support those of Reed et al,17 who compared outcomes among children hospitalized with S aureus. Children coinfected with influenza were more likely to experience severe outcomes, including intensive care unit admission and death, than children with either influenza or S aureus infection alone.17 Given the relatively small number of patients with S pneumoniae and influenza coinfection, this portion of the subanalysis was likely underpowered to detect meaningful differences in outcomes.
This study has several limitations. Discharge diagnosis codes may be unreliable for specific diseases or pathogens. We attempted to minimize the likelihood of misclassification of patients with complicated pneumonia by using restrictive inclusion criteria. The study cohort was limited to patients with discharge diagnoses of pneumonia and pleural effusion who received antibiotic therapy and a pleural fluid drainage procedure. Any misclassification of patients with complicated pneumonia, however, would likely be nondifferential, biasing our results toward finding no difference between children with and without influenza coinfection. Misclassification of bacterial pathogens, while possible, was also likely minimal because the microbiological yield was consistent with prior studies of complicated pneumonia5,7,8,37,38 and nondifferential because the proportion of patients with a specified bacterial cause was similar between the patients with and without influenza coinfection. Last, some patients with influenza coinfection may not have been identified as a consequence of miscoding or absence of influenza testing. Misclassification as a result of errors in coding is less likely as antiviral medications were infrequently administered to those classified as not having influenza coinfection. However, if influenza testing were performed disproportionately in children with severe illness, a failure to identify some coinfected children with mild disease would result in an overestimation of the consequences of coinfection. In a subanalysis restricted to those tested for influenza, outcomes remained consistently worse for coinfected children, suggesting that spectrum bias alone does not account for our study findings. Additionally, the prevalence of influenza coinfection in our study is similar to previous reports of children with primarily uncomplicated pneumonia,16- 18 and the finding of increased severity among coinfected individuals is supported by several recent studies.17,18
Among children hospitalized with complicated pneumonia, those coinfected with influenza had increased disease severity. These findings are consistent with other studies performed on a smaller scale and serve to alert the clinician that viral testing is an important consideration in children with complicated pneumonia, particularly in those with a more severe clinical course. Moreover, our findings underscore the importance of routine influenza vaccination for children.
Correspondence: Samir S. Shah, MD, MSCE, Room 1526 (North Campus), Division of Infectious Diseases, Children's Hospital of Philadelphia, 34th Street and Civic Center Boulevard, Philadelphia, PA 19104 (firstname.lastname@example.org).
Accepted for Publication: November 24, 2010.
Published Online: February 7, 2011. doi:10.1001/archpediatrics.2010.295
Author Contributions:Study concept and design: Williams, Hall, Brogan, Myers, Newland, and Shah. Acquisition of data: Williams and Hall. Analysis and interpretation of data: Williams, Hall, Brogan, Farris, Myers, and Shah. Drafting of the manuscript: Williams, Hall, Brogan, Myers, and Shah. Critical revision of the manuscript for important intellectual content: Williams, Brogan, Farris, Myers, Newland, and Shah. Statistical analysis: Hall and Shah. Obtained funding: Shah. Administrative, technical, and material support: Brogan and Myers. Study supervision: Williams and Shah.
Financial Disclosure: None reported.
Funding/Support: Dr Shah was supported by grant K01 AI73729 from the National Institute of Allergy and Infectious Diseases and the Robert Wood Johnson Foundation under its Physician Faculty Scholar program.
Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Williams DJ, Hall M, Brogan TV, Farris RWD, Myers AL, Newland JG, Shah SS. Influenza Coinfection and Outcomes in Children With Complicated Pneumonia. Arch Pediatr Adolesc Med. 2011;165(6):506-512. doi:10.1001/archpediatrics.2010.295