Factors Associated With Antibiotic Use for Children Hospitalized With Pneumonia

This was a secondary analysis of a prospective cohort study (Catalyzing Ambulatory Research in Pneumonia Etiology and Diagnostic Innovations in Emergency Medicine [CARPE DIEM]) of children with suspected CAP who presented to a tertiary-care pediatric emergency department (ED).^13–15  The study was approved by the hospital’s institutional review board. Informed consent was obtained from all legal guardians and assent was obtained from children ≥11 years of age.

Children 3 months to 18 years of age with signs and symptoms of a lower respiratory tract infection who received a CXR in the ED for suspected CAP between July 2013 and December 2017 were eligible to be enrolled. Signs and symptoms of a lower respiratory tract infection included 1 or more of the following: cough, sputum, chest pain, shortness of breath, tachypnea, or abnormal lung findings (eg, crackles, wheezing) on physical examination.^5,14  We did not include fever in the inclusion criteria, similar to previous pneumonia literature,¹¹  because both fever in the ED and fever at home have limitations; children may have already received an antipyretic before presentation, and parents may have different definitions of fever and methods for checking for one.⁵  To be inclusive of all patients with suspected CAP, children who met all inclusion criteria and may have had a potential asthma exacerbation were not specifically excluded. Because the study was intended to investigate CAP in otherwise healthy children, patients were excluded if they had complex chronic conditions (eg, congenital heart disease, tracheostomy-dependent, chronic lung disease, neuromuscular disease, sickle cell disease, immunodeficiency), history of aspiration pneumonia,¹⁶  hospitalization within the previous 14 days, or previous study enrollment within the last 30 days. Additionally, for this analysis, children who received >2 days of antibiotics before the ED visit and those discharged from the ED were excluded. Thus, this analysis only included children who presented to the ED and were ultimately hospitalized, including patients admitted to the inpatient ward or ICU.

Patient demographics, medical history, current symptoms, and ED clinicians’ physical examination findings were prospectively collected by trained research coordinators. Imaging results, antibiotic therapy, and medical interventions were extracted from the electronic health record and reviewed for accuracy by 2 investigators. Respiratory viral testing was not included in this analysis because testing as a part of clinical care was uncommon and previous studies have revealed that viral testing was not associated with antibiotic decisions.¹⁷ 

The primary outcome was the receipt of inpatient antibiotics, defined as at least 1 dose of antibiotics at any time during the hospitalization, after transfer out of the ED. Our goal was to evaluate the risk factors associated with any inpatient antibiotic use, regardless of the number of doses because previous literature revealed that each additional day of antibiotics was associated with measurable antibiotic harm.¹⁸  Because some patients may receive just 1 dose of antibiotics in the inpatient setting before discontinuation, our secondary outcome was the receipt of a full course of antibiotics, defined as ≥5 days of antibiotics¹⁹  during the admission or antibiotic prescription at discharge.

All CXRs were interpreted based on a pediatric radiologist’s impression as part of clinical care and manually categorized into 1 of the 4 mutually exclusive groups: (1) radiographic CAP (findings such as “infiltrate,” “consolidation,” or “pneumonia”), (2) equivocal (“atelectasis vs pneumonia”), (3) atelectasis (“atelectasis” without mention of the terms that meet the criteria for radiographic CAP), and (4) normal (CXRs that did not meet the criteria for previous categories and included findings such as “normal,” “peribronchial thickening,” or “airway disease”). We focused on these 4 categories because there are many abnormal CXR findings, and we aimed to evaluate the association between antibiotic use and CXR findings across the spectrum of abnormal results. Additionally, previous studies also separated equivocal CXR findings and radiographic CAP.^12,20  Similarly to the authors of previous literature, we defined nonradiographic CAP as any CXR that did not meet the criteria for radiographic CAP or equivocal CXR (ie, normal CXRs or those with atelectasis).¹⁵ 

Descriptive statistics were computed overall and by inpatient antibiotic use groups. Continuous and categorical variables were compared by using the Wilcoxon rank test and Pearson’s χ² test, respectively. We performed univariable and multivariable Poisson regression with robust error variance to determine the association between risk factors and inpatient antibiotic use.²¹  This approach allows us to estimate the relative risk (RR) directly, which is recommended over odds ratios for prospective studies with a nonrare binary outcome.^21,22  In multivariable models, we adjusted for risk factors identified in 2 ways: (1) those determined a priori to be clinically relevant, including age, history of fever, history of asthma (including reactive airways disease), receipt of at least 1 dose of antibiotics in the ED, CXR findings, and severe illness in the ED and (2) those with P < .01 in bivariable analysis (ie, days of illness, wheeze, focal crackles, and focal decreased breath sounds in the ED, and receipt of supplemental oxygen in the inpatient setting). Severe illness in the ED was defined as receipt of supplemental oxygen, positive-pressure ventilation, or vasopressors in the ED, diagnosis of sepsis,²³  or transfer directly from the ED to the ICU.^24,25  An interaction term was included to assess how the receipt of antibiotics in the ED modifies the effect of CXR findings on inpatient antibiotic prescribing. This was decided a priori as ED antibiotic decisions may differentially affect how CXR findings influence inpatient antibiotic prescribing. We performed a subgroup analysis to examine the risk factors among children with nonradiographic CAP because they are at low risk for bacterial disease.¹¹  Statistical tests were 2-sided with a significance level of 0.05. Analyses were performed by using SAS (version 9.4, Cary, NC).

Of 1142 children enrolled, 477 met the inclusion criteria for the current study (Supplemental Fig 1). The median age was 2.8 years (interquartile range, 1.2–5.8). Most children reported having fever (86%) and cough or rhinorrhea (82%) (Table 1). Half of the patients (51%) had radiographic CAP or equivocal CXR and 49% had nonradiographic CAP.

In total, 285 (60%) patients received at least 1 dose of antibiotics in the inpatient setting, and 254 (53%) received a full antibiotic course. Children who received antibiotics in the inpatient setting were older (3.3 vs 2.2 years), had a longer fever duration (3 vs 2 days), and more often had radiographic CAP (47% vs 3%) and abnormal lung findings in the ED (Table 1).

Of the children with radiographic CAP, 96% received inpatient antibiotics and 91% received a full antibiotic course. Of the 232 children with nonradiographic CAP, 29% received inpatient antibiotics and 21% received a full course. Children with nonradiographic CAP represented 24% of all children who were given inpatient antibiotics and 21% who were given a full antibiotic course.

Most patients who received antibiotics in the ED also received them in the inpatient setting (90%), and most children who received antibiotics in the inpatient setting went on to receive a full course (89%). For those with nonradiographic CAP, most who received antibiotics in the ED had antibiotics continued in the inpatient setting (71%), and most who received antibiotics in the inpatient setting completed a full course (72%).

In adjusted analyses, receipt of antibiotics in the ED, history of fever, and receipt of supplemental oxygen in the inpatient setting were associated with an increased risk of inpatient antibiotic use (Table 2; Supplemental Table 4 for unadjusted analyses). The association between CXR findings and inpatient antibiotics was modified by the receipt of antibiotics in the ED (P < .001; Table 3). Children with radiographic CAP and equivocal CXR had an increased risk of receiving inpatient antibiotics compared with those with normal CXRs, but the increased risk was modest when antibiotics were given in the ED (RR 1.35 [1.06–1.71] and 1.36 [1.06–1.74], respectively) and much greater when ED antibiotics were not given (RR 6.07 [3.86–9.54] and 3.67 [2.09–6.44], respectively). Additionally, when antibiotics were given in the ED, there was no difference in the risk of inpatient antibiotic use between those with equivocal and radiographic CAP. Among children who did not receive ED antibiotics, those with atelectasis had an increased risk of receiving inpatient antibiotics compared with those with a normal CXR (RR 2.03 [1.09–3.76]).

There were 232 children with nonradiographic CAP, and in the adjusted model, those who received antibiotics in the ED had an increased risk (RR 2.83 [1.87–4.28]) of receiving inpatient antibiotics (Supplemental Tables 5–6). History of fever, focal crackles, and receipt of supplemental oxygen in the inpatient setting also increased the risk of inpatient antibiotic use in this cohort.

We found similar risk factors for a full course of antibiotics as we did in our primary analysis (Supplemental Tables 6–7).

In this prospective cohort study of children hospitalized with suspected CAP, antibiotic use in the ED, abnormal CXR findings, fever, and the use of supplemental oxygen were associated with the receipt of antibiotics in the inpatient setting and a full treatment course. Antibiotics initiated in the ED were frequently continued in the inpatient setting (90%), and those who received a dose in the inpatient setting often completed a full course (89%). Despite evidence that most children with nonradiographic CAP are at low risk for bacterial CAP,¹¹  nearly 1 in 3 received inpatient antibiotics and 1 in 4 received a full treatment course. In this subgroup, inpatient antibiotic use was associated with antibiotics in the ED, fever, focal crackles, and the use of supplemental oxygen.

Although antibiotics are not beneficial for most children with nonradiographic CAP,¹¹  21% of those who received a full antibiotic course had nonradiographic CAP. Our findings are supported by previous studies revealing that 22% to 36% of children in the ED who received antibiotics had nonradiographic CAP.^11,12  Although clinicians may be concerned that CXR findings consistent with pneumonia can be absent early on or in patients with dehydration, there are no pediatric data to support this, and a previous study revealed the high (98%) negative predictive value of CXRs.^11,26,27  Although CXRs are imperfect (particularly their inability to differentiate viral versus bacterial and infectious vs noninfectious causes when infiltrates are present), clinicians can harness their negative predictive value to identify those who likely have a nonbacterial respiratory illness (eg, viral pneumonia, bronchiolitis, asthma) and would not benefit from antibiotics.^11,28,29  Our work highlights the great potential for reducing unnecessary antibiotics by targeting children with nonradiographic CAP.

When evaluating the impact of CXRs across the spectrum of findings, we found that children with atelectasis who did not receive antibiotics in the ED had more than a 2-fold increased risk of receiving inpatient antibiotics and a full treatment course compared with those with normal CXRs. Few studies have explored the association between antibiotic use and atelectasis. Our finding suggests that, despite atelectasis being a nonspecific sign on many pediatric CXRs, atelectasis increased the risk of antibiotic use.³⁰  Additionally, among children who received antibiotics in the ED, there was no difference in the risk of inpatient antibiotics between those with equivocal CXRs or radiographic CAP, although the term “equivocal” suggests uncertainty regarding the presence of radiographic CAP. Similarly, previous studies reveal high antibiotic rates for equivocal CXRs.¹²  Although we acknowledge the subjectivity of CXR interpretations,³¹  our results suggest that optimizing decision making in the context of nuanced CXR findings may help promote antibiotic stewardship.

Most children who received antibiotics in the ED had antibiotics continued in the inpatient setting and received a full treatment course, even among those with nonradiographic CAP. Additionally, when antibiotics were given in the ED, children with abnormal CXR findings had only a modest increased risk of inpatient antibiotic use compared with those with normal CXRs. The inpatient setting provides a unique opportunity for clinicians to make therapeutic decisions based on an evolving clinical picture rather than 1 point in time.³²  However, our data suggest that inpatient antibiotic decisions were rarely modified and were strongly influenced by ED treatment decisions. A single-center study of 181 children revealed only 38% of those who received antibiotics in the ED continued antibiotics in the inpatient setting.³³  These results may differ from ours because of varied hospital practices or patient populations because our study included children with suspected CAP who had a CXR, whereas the other study included a higher proportion (54%) of children with asthma and 19% without a CXR. Similar to our findings, a scenario-based survey of adults revealed that hospitalists were more likely to continue both inappropriate and overly broad antibiotic therapy when initiated by ED clinicians.³⁴ 

The influence of ED antibiotic prescribing on inpatient treatment decisions may relate to the concept of therapeutic momentum (a term used, at times, synonymously with clinical inertia and therapeutic inertia).^35–39  Therapeutic momentum is the failure of clinicians to initiate or intensify therapy when therapeutic goals are not reached or to stop or reduce therapy that is no longer needed.^35–39  Based on previous literature, there exists a strong norm for noninterference, in which clinicians do whatever they can to avoid altering another prescriber’s decisions, and this may play a role in therapeutic momentum in pediatric CAP.⁴⁰  The desire to continue treatment of a patient that is improving and overestimate the potential benefits of antibiotics, particularly in the context of simultaneously receiving many other ED interventions (eg, intravenous fluids) that improve a child’s clinical appearance, may also be a contributing factor.⁴¹  Additionally, clinician perceptions of parental pressure or expectations with regard to antibiotics, similar to its role in antibiotic decision making in the ambulatory setting,⁴²  may contribute in the inpatient setting, particularly once antibiotics have already been started. Therapeutic momentum is a recognized barrier to the appropriate escalation or deescalation of medications for adults and children with chronic conditions such as diabetes and hypertension.^43,44  Less is known about therapeutic momentum in the acute care setting because patients transition from the ED to inpatient wards or its role as a modifiable barrier to antibiotic stewardship, making our study a novel contribution with important implications.

Our study has several limitations. First, this study was conducted at a single tertiary care academic pediatric hospital, and conclusions, particularly about the influence of ED practices on antibiotic prescribing in the inpatient setting, may not be generalizable to other settings. Second, because we only included patients in whom a CXR was obtained, the applicability of these findings for children without CXRs is unknown. However, given that 89% of patients admitted with CAP at children’s hospitals in the United States receive CXRs, we suspect this will be applicable to many institutions.^4,45  Third, we may have misclassified patients who also had an asthma exacerbation or reactive airway disease, which may have been disproportionally higher for those with nonradiographic CAP. This may have influenced our antibiotic risk factors, particularly for patients with nonradiographic CAP. However, children included in the study had to have a CXR performed for suspected CAP regardless of additional suspicion for asthma. In addition, our analysis adjusted for a history of asthma or reactive airway disease to reduce this confounder. Fourth, CXR interpretation is observer-dependent.^2,31  The aim of this study was to understand how factors in clinical practice, such as CXR results presented in real time, influence antibiotic decision making; thus, we categorized CXR findings on the basis of the interpretation made in clinical practice. Fifth, the original prospective cohort study was powered to differentiate severity in children with CAP,²⁴  and because this was a secondary analysis, we report our findings as effect sizes. Finally, because we do not know the indication for antibiotic use, we cannot be certain antibiotics were prescribed for CAP rather than other suspected bacterial infections.

Among children hospitalized with suspected CAP, receipt of antibiotics in the ED, CXR findings, history of fever, and receipt of supplemental oxygen were strongly associated with prescribing antibiotics in the inpatient setting and receiving a full treatment course. Nearly a third of children with nonradiographic CAP received inpatient antibiotics despite having a low risk for bacterial CAP. Once antibiotics were started in the ED, they were often continued, even for patients with nonradiographic CAP. Therapeutic momentum, rather than a clear rationale for continuing antibiotics, likely played a role in these findings. Targeting therapeutic momentum and implementing other antibiotic stewardship strategies, focusing particularly on children with nonradiographic CAP, could help improve judicious antibiotic use.

We acknowledge Jessi Lipscomb and Judd Jacobs for their role in data management for the CARPE DIEM study. Mantosh Rattan, MD, and Eric Crotty, MD, from the Department of Radiology at Cincinnati Children’s Hospital Medical Center reviewed and interpreted all chest radiographs. We would also like to acknowledge Caitlin Clohessy and Andrea Kachelmeyer for recruiting patients and performing chart review to verify the collected data. We are grateful to the entire research team and patient services staff in the Divisions of Emergency Medicine and Hospital Medicine at CCHMC for their assistance with study procedures. Finally, we are especially grateful to the patients and families who enrolled in the CARPE DIEM study.

CAP

community acquired pneumonia

CXR

chest radiograph

ED

emergency department

RR

relative risk