Associations Between Weight and Lower Respiratory Tract Disease Outcomes in Hospitalized Children Free

We performed a retrospective cohort study using data from the PEDSnet clinical research network (https://pedsnet.org/). PEDSnet institutions share electronic health record and other clinical data to rapidly accelerate pediatric evidence generation. The PEDSnet data from this study came from 6 major children’s health systems across the United States. The different health systems vary in size, with the smallest contributing 1.6% and the largest 26.4% of hospital encounters for this cohort. This study was determined to qualify for exemption from human subjects review by our local institutional review board.

We identified children ages 2 to 18 years evaluated in the emergency department and subsequently hospitalized at participating sites between January 1, 2009 and November 30, 2019 with any diagnosis code of bronchiolitis, viral pneumonia, asthma, or CAP. Although we did not anticipate many children in this age group would be diagnosed with bronchiolitis, we included the diagnosis to optimize our capture of children with respiratory illness and to be consistent with previous literature.^20,21  To identify the diagnoses of interest, we started with previously published literature of International Classification of Diseases, 9^th Revision, Clinical Modification codes (Supplemental Table 3),^20,21  which were translated to the appropriate Systematized Nomenclature of Medicine Clinical Terms codes (Supplemental Table 4). Since only Systematized Nomenclature of Medicine Clinical Terms codes were used to query the PEDSnet database, the transition to International Classification of Diseases, 10^th Revision, Clinical Modification coding did not affect our search strategy. We excluded children with complex chronic conditions defined by the Pediatric Medical Complexity Algorithm Version 3.0, because these children are expected to have different growth patterns than children without medical complexity.²²  Asthma and obesity are diagnoses included in this algorithm but were not used to exclude children for this study. We also excluded children without documented BMI within 3 months of the index encounter. Overweight and obesity are not consistently defined in children <2 years, and patients >18 years are not consistently admitted to children’s hospitals; therefore, these children were excluded.^23,24 

Data obtained from the PEDSnet database included dates of encounters, anthropometrics, diagnoses, demographics, and procedure codes. Payer was categorized as public, private or military, or none. Some encounters had multiple payers listed; in that case, the hierarchical order chosen was private or military (if available), followed by public (if available), followed by none. Diagnoses were categorized as asthma, bronchiolitis, CAP, and viral pneumonia and flagged if listed as the primary diagnosis for that encounter. For the primary exposure, we used the height and weight from the index emergency department (ED) visit to compute BMI. If either height or weight or both were not recorded during the index ED visit, we selected the closest available recorded weight ± 90 days of the index encounter followed by any height obtained ± 60 days of the weight measurement to minimize selection bias. We used age on the date of the encounter, sex, and BMI to classify the child’s weight per Centers for Disease Control and Prevention guidelines as follows: <5%ile underweight, 5%ile to <85%ile normal weight, 85%ile to <95%ile overweight, 100% to <120% of 95%ile class 1 obesity, 120% to <140% of 95%ile class 2 obesity, and >140% of 95%ile class 3 obesity.^7,23,24  Outliers were identified by CDC guidelines and excluded.²⁵  Age was categorized as 2 to 5 years, 6 to 11 years, and 12 to 18 years.

Our outcome measures assessed severity of illness, measured by receipt of positive pressure ventilation (continuous positive airway pressure, bilevel positive airway pressure, or mechanical ventilation) during the hospital stay, ICU admission at any point during the hospitalization, and hospital LOS. We were unable to assess nasal cannula or high-flow nasal cannula use in this data set. In addition, we assessed 7- and 30- day ED revisits and 7- and 30-day hospital readmissions for asthma, bronchiolitis, CAP, or viral pneumonia.

We defined an asthma subgroup for this analysis, which included all patients with a primary diagnosis of asthma. We expected those patients to be managed primarily for an asthma exacerbation during their hospital admission, which was corroborated by review of medications (specifically bronchodilators) administered during the hospitalization. In addition, we defined subgroups for CAP and for viral pneumonia and bronchiolitis, which included children with those listed as primary or secondary diagnoses.

Results were analyzed initially using descriptive statistics to determine the prevalence of overweight and obesity among children by demographic and clinical characteristics, including age, sex, race and ethnicity, payer, and respiratory diagnosis. Bivariate comparisons between groups were done by using χ² tests for proportions and analysis of variance procedures for continuous variables. Along with our primary exposure of BMI category, we included age, sex, and payer in a mixed effects multivariable logistic regression model, which treated site, patient, and year of admission as random effects. The outcome of LOS was modeled using Poisson regression, both unadjusted and adjusted for age category, sex, and payer. A 2-sided P value of .05 was used as the cutoff for statistical significance. Sensitivity analyses were performed with all payer data from site D set to missing, given known loss of data at that site due to electronic health record migration during the study period. Additionally, sensitivity analyses were performed by using only data from sites C, F, and G, which had the lowest percentages of missing BMI data. All analyses were performed at the encounter level using SAS version 9.4 (SAS Institute, Cary, NC).

We identified 47 693 children who met our inclusion and exclusion criteria, as shown in Fig 1; of those, 6.7% had underweight, 57.8% normal weight, 14.6% overweight, 13.2% class 1 obesity, 5.0% class 2 obesity, and 2.8% class 3 obesity (obesity class groupings are exclusive of each other). These patients had a total of 65 132 hospitalizations during the study period; demographic and clinical information for all encounters in this cohort are shown in Table 1. The distribution of children with any diagnosis of CAP (P < .01), any diagnosis of viral pneumonia or bronchiolitis (P < .01), and a primary diagnosis of asthma (P < .01) differed between weight categories. Clinical and demographic information for the subgroup of children with primary asthma are shown in Supplemental Table 5.

Our primary outcome was severity of illness, measured by need for positive pressure ventilation and ICU care. Figure 2 demonstrates associations between BMI category and positive pressure ventilation following mixed effects multivariable logistic regression. Class 3 obesity had the strongest associations in both the full cohort (odds ratio [OR] 1.62, 95% confidence interval [CI] 1.38–1.89) and in the subgroup with a primary diagnosis of asthma (OR 1.65, 95% CI 1.33–2.06); no association was seen with class 2 or 3 obesity in the CAP or viral pneumonia or bronchiolitis subgroups. Statistically significant associations were seen for overweight and class 1 obesity in all four cohorts. Underweight had a positive association in the full cohort (OR 1.39, 95% CI 1.24–1.56) and the viral pneumonia or bronchiolitis subgroup (OR 1.19, 95% CI 1.00–1.41), but no associations were seen in the asthma or CAP subgroups.

Figure 3 demonstrates associations between BMI category and ICU care. For this outcome, the strongest associations were found with underweight for both the full cohort (OR 1.40, 95% CI 1.30–1.52) and the asthma subgroup (OR 1.41, 95% CI 1.26–1.59). In contrast, the strongest associations were found with class 3 obesity for the CAP subgroup (OR 1.82, 95% CI 1.05–3.13) and the viral pneumonia or bronchiolitis subgroup (OR 1.68, 95% CI 1.00–2.81). As seen above with respiratory support, overweight and class 1 obesity had statistically significant associations across all four cohorts. In addition, we performed sensitivity analyses to verify that known issues with insurance data from 1 site as well as the degree of missing BMI data overall did not impact our results; minimal differences in the multivariable models were noted in those sensitivity analyses (Supplemental Tables 5 and 6).

Associations with hospital LOS are shown in Table 2. After both unadjusted and adjusted analyses, the strongest associations in the full cohort were seen with underweight, with coefficients 0.27 (95% CI 0.25–0.29) and 0.32 (95% CI 0.30–0.33), respectively. Differences in LOS between other weight categories, which range from 2.6 to 2.9 days, seem unlikely to be clinically significant even though some statistically significant associations were found. In each of the subgroup analyses, stronger positive associations were seen with obesity rather than with underweight.

Although 7- and 30-day ED revisit and hospital readmission prevalence was higher for children with overweight or obesity, these findings remained consistent only for the 30-day outcomes in the multivariable regression models, as shown in Supplemental Table 7.

Using data from 6 geographically diverse children’s health systems, we found that children with underweight and those with overweight or obesity were more likely to receive positive pressure respiratory support and ICU admission when hospitalized with asthma or LRTI. In the full cohort, underweight, but not overweight or obesity, was associated with a longer hospital LOS, whereas obesity was associated with an increase in 30 day ED revisits and hospital readmissions. To our knowledge, this is the first large multicenter study in the United States to evaluate these associations using measured anthropometrics.

We found that underweight, overweight, and obesity were associated with positive pressure respiratory support, with a dose-dependent increase with increasing obesity. This remained relatively consistent across our subgroup analyses. This finding corroborates a previous study using the Kids Inpatient Database, in which obesity was associated with the use of mechanical ventilation in hospitalized children with a diagnosis of bronchitis or pneumonia (adjusted OR 2.90, 95% CI 2.15–3.90).¹³  In that study, children with obesity were identified solely by administrative coding with cohort prevalence of obesity of 1.1%; the larger effect size seen in that study could be explained by clinicians being more likely to diagnose obesity in children with more severe obesity, in whom stronger associations with this outcome might be expected. In contrast, one previous small study of children admitted with influenza-like illness did not find significant associations between overweight or obesity and prolonged LOS or oxygen requirement; however, the study may have been underpowered to detect these outcomes, as it enrolled only 134 hospitalized children.¹⁹  Pediatric obesity has previously been associated with risk of respiratory failure in a variety of settings, including sedations and operations.^26–29  Despite this literature, child weight status was not included as a potential predictor in a recent risk prediction model for pneumonia severity of illness; it should be considered to improve risk stratification in the future.³⁰  Awareness of both underweight and obesity as risk factors for severe disease could help clinicians making decisions about patient triage, monitoring, and disposition.

For the outcomes of ICU admission and hospital LOS, we found stronger associations with underweight compared to overweight or obesity. These findings could reflect confounding by malnutrition, recent illness, or complex conditions not captured in the Pediatric Medical Complexity Algorithm used, although our study was not designed specifically to investigate these mechanisms. Similar associations between underweight and disease severity have been found previously in children hospitalized in Japan for influenza-related respiratory infections and respiratory syncytial virus after excluding patients with medical complexity.^31,32  Underweight could be associated with malnutrition or micronutrient deficiencies; vitamin D deficiency may be associated with poor outcomes from sepsis and respiratory illness in children, and zinc supplementation may reduce risk of mortality from pneumonia.^33,34  Further research is necessary to understand the varied mechanisms behind these associations and mitigate poor outcomes in this population.

The existing literature on pediatric obesity and asthma is inconsistent; therefore, we performed a subgroup analysis of children with a primary diagnosis of asthma to further investigate these associations. Our findings in this cohort were similar to the overall cohort, specifically an increased need for positive pressure respiratory support and ICU admission with increasing weight category; additionally, we found a positive association between LOS and increasing weight category in this cohort. Multiple single-site studies have produced contradictory results on associations between obesity and LOS or ICU admission^9–11,35  One national study using the Kids’ Inpatient Database relied on administrative codes for the definition of obesity and did find associations with mechanical ventilation and longer hospital length of stay for children admitted with asthma.¹⁴  Our study validates this finding with the use of measured height and weight. In contrast, a single-center study conducted over 1 year found that obesity had minimal association with LOS or ICU admission in children hospitalized with status asthmaticus.⁹  Our multiyear, multicenter study represents a significant improvement in terms of statistical power and generalizability compared with this existing literature. It has been suggested that pediatric obesity-related asthma is a distinct asthma phenotype with increased morbidity and impaired response to common asthma medications^36,37 ; in our study, we did see increased reliance on advanced respiratory support in this population, but we did not specifically assess medication response. Our findings suggest that clinicians should consider a patient’s weight when triaging children presenting with respiratory illness, especially given known underrecognition of obesity in the inpatient setting.^15–17  Future research on health outcome disparities and underlying mechanisms directing health outcomes for children with underweight, overweight, or obesity could inform evidence-based guidelines specific to the inpatient management of these populations.

This study used measured anthropometric data from multiple sites across the United States to investigate associations between weight and respiratory disease outcomes. The prevalence of obesity in our study is very similar to the prevalence of 20.6% previously noted for children hospitalized for infectious causes, providing further reassurance about the generalizability of our findings.³⁸  However, our study should be interpreted in light of several limitations. All sites included in the PEDSnet database are children’s hospitals, so results may not be generalizable to all institutions providing care for children. A significant proportion of all potential encounters in the PEDSnet database had missing height data that prevented the calculation of BMI; it is possible that exclusion of these subjects could bias our analysis, although our sensitivity analysis using data from the three sites with the least missing data are reassuring. Even with measured height and weight, BMI is an imperfect measure of adiposity and may not accurately represent body composition for all patients.^39,40  Our cohort was defined using diagnosis codes for respiratory disease, which may result in misclassification of some patients. Most notably, changes in the designation of a primary diagnosis code over time may have complicated our subgroup analysis of children with a primary diagnosis of asthma.⁴¹  Furthermore, associations between weight categorization and clinical outcomes could vary with different specific infections (ie, coronavirus disease 2019, pandemic influenza A [H1N1]), which we did not assess in this study.^42–44  The cohort was defined using children initially seen in an ED, thus excluding those directly admitted to the hospital. Despite the use of electronic health record data in the PEDSnet data set, there were some variables of interest for which we were unable to obtain granular data (ie, supplemental oxygen via nasal cannula). For children who did require positive pressure ventilation, we have limited information about the reasons for their respiratory failure. ICU admission was another primary outcome; however, indications for ICU admission could vary between institutions. Socioeconomic status could be a significant confounder in the association between weight status and the outcomes assessed in this study. Payer was the only available proxy but may not be an ideal measure of socioeconomic status, which could contribute to residual confounding; our study was not well-positioned to evaluate potential associations between social determinants of health and hospital outcomes. Finally, even in this large, multisite cohort, a relatively small number of children with class 2 and 3 obesity may have limited our power to identify associations in these populations.

Among children evaluated for lower respiratory tract disease, both underweight and overweight or obesity are associated with increased respiratory support and ICU admission in hospitalized children. Further research is needed to understand these associations and their underlying mechanisms, with the goal of improving clinical outcomes for all children.