a b s t r a c t

Background: Acute asthma exacerbations (AAE) account for many Pediatric Emergency Department (PED) visits. Chest radiography (CXR) is often performed in these patients to identify practice-changing findings such as pneumonia (PNA). Limited knowledge exists to balance the cost and radiation dose of CXR with expected yield of Clinically meaningful information.

Objective: To determine in children with AAE with CXR, whether patient characteristics are associated with radiographic PNA; and significant practice change by initiation of antibiotic.

Design/Methods: Retrospective chart review of AAE patients with CXR performed in a PED in 2014. We examined univariate associations between patient characteristics and PNA on CXR and administration of antibiotic. Multi- ple logistic regression models then subsequently examined adjusted associations between patient characteristics and both outcomes.

Results: Of 288 patients, 43 (15%) had PNA on CXR and 51 (17.8%) received antibiotics. There were no statistically significant univariate associations between either outcome and age, race, gender, insurance status, mode of PED arrival, fever or hypoxia (all p N 0.11). Crackles were associated with antibiotic administration (p = 0.03), but not PNA on CXR (p = 0.07). Only previous antibiotic use within 7 days had both significant univariate associations (p = 0.002) and adjusted associations with both PNA on CXR (aOR 3.6) and antibiotic administration (aOR 3.3). Conclusion: CXR infrequently adds valuable information in children with AAE. Patients treated with antibiotic within 7 days are more likely to have PNA identified on CXR and receive antibiotics. A larger study is needed to examine potential significance of hypoxia and crackles.

(C) 2017

Introduction

Pediatric acute asthma exacerbations (AAE) are responsible for ap- proximately 640,000 pediatric emergency department (PED) visits each year and result in substantial Healthcare costs [1]. Chest radiogra- phy (CXR) remains the most frequently ordered ancillary test with high variability among providers despite recommendations that it should not be a routine part of the evaluation of pediatric acute asthma exacerbation [2-4].

Abbreviations: AAE, Acute Asthma Exacerbation; PED, Pediatric Emergency Department; CXR, Chest Radiograph; PNA, Pneumonia.

* Corresponding author.

E-mail address: [email protected] (E.H. Allie).

It is widely accepted that CXR should be obtained only in those pa- tients with localized rales or dullness that do not resolve after adequate bronchodilator treatment (to evaluate for pneumonia); those with oxy- gen saturation b 88% (to evaluate for pneumothorax); and those with significant chest pain (to evaluate for pneumothorax and pneumomediastinum) [3,4]. Despite this recommendation, between 27% to 43% of children with asthma exacerbation receive a CXR in the emergency department setting [5] and fewer than 5% of CXRs in pediat- ric asthma patients have treatment-altering findings such as pneumonia [2]. Examination findings such as fever, hypoxia and localized areas of dullness on pulmonary examination can be present in patients with pneumonia (PNA) yet are often present in patients with AAE who do not have PNA. As such, the presence of these findings may prompt pre- mature clinician ordering of CXR during initial Patient evaluation. In- deed, N 80% of childhood asthma exacerbations are precipitated by

https://doi.org/10.1016/j.ajem.2017.10.009

0735-6757/(C) 2017

viral Respiratory infections, many of which result in fever and airway mucous plugging and atelectasis may result in localized dullness or rales on physical examination [6,7] . Performing CXR in this great major- ity of children with Acute asthma exacerbations, those in whom the ex- acerbation is precipitated by a viral respiratory infection, may not be in the patient’s best interest. In addition to the unnecessary radiation dose and cost of the study, atelectasis is frequent during exacerbations and may be difficult to distinguish from infiltrate due to pneumonia. As a re- sult, patients may receive inappropriate antibiotic treatment.

Investigations of pediatric patients with acute asthma exacerbations have examined associations of abnormal CXR findings, in particular pneumonia, with demographic and clinical patient characteristics [2]. However, there is limited knowledge of patient characteristics associat- ed with CXR receipt and of identified CXR abnormalities that change pa- tient management. Abnormalities which may prompt a clinical intervention that significantly deviates from usual care of AAE include pneumomediastinum, pneumothorax and pneumonia (lobar/bacterial)

[3] with concomitant pneumonia being identified most frequently of the three. The discovery of spontaneous pneumomediastinum or pneu- mothorax in pediatric patients rarely warrants a significant clinical in- tervention such as thoracostomy tube placement or other surgical intervention, and the existence of AAE with concomitant pneumonia on CXR is far more common [3,8-10]. Our objectives in conducting this study were twofold. First, we sought to examine whether characteristics of children with AAE in a PED are associated with pneumonia on CXR when this study is obtained. Second, we sought to examine whether CXR findings are associated with a significant change in clinical manage- ment from typical asthma exacerbation therapy in the initiation of anti- biotic treatment for pneumonia.

Methods

We performed a retrospective chart review of pediatric patients with acute asthma exacerbations in an urban, tertiary children’s hospi- tal PED with approximately 55,000 visits per year. Inclusion criteria were age 36 months to 17 years of age, previously documented history of asthma, presentation with an exacerbation between January 1, 2014 to December 31, 2014 and receipt of CXR as part of diagnostic evalua- tion. Eligible patients were identified using the International Classifica- tion of Diseases, Ninth Revision (ICD-9) diagnosis code for asthma (ICD 493.90-493.93). Exclusion criteria included age b 36 months, age N 17 years, trauma, and presence of tracheostomy, respiratory failure with endotracheal intubation, chronic lung disease, and sickle cell ane- mia, primary complaint of chest pain or anxiety, or complex past medi- cal history. Children b 36 months were excluded due to concern for similarities in the presentation of acute bronchiolitis and asthma exac- erbation without a formal asthma diagnosis [11].

Data were extracted and reviewed from our electronic medical re- cord and included available demographic and clinical variables: age, gender, race, insurance status, mode of arrival to the PED, pre- treatment oxygen saturation on room air by pulse oximetry, presence of absence of documented fever of at least 38 ?C, treatment with antibi- otic within 7 days prior to PED visit, presence of crackles on physical ex- amination, findings of radiographically significant pneumonia on chest radiograph, and receipt of antibiotic in the PED as a result of CXR find- ings. All data points were recorded by three clinical physicians and one medical student with medical student chart extraction re- reviewed by principal investigator to ensure accurate data extraction.

Radiographically-significant pneumonia was defined as focal consol- idation in a typical or atypical pattern as determined by the reading pe- diatric radiologist. Chest radiographs that were interpreted by a pediatric radiologist as unequivocally positive for radiographically- significant pneumonia were designated ‘positive’ for analyses. CXRs interpreted as ‘indeterminate’ by the initial pediatric radiologist were reviewed independently by two pediatric radiologists. Chest radio- graphs for which these radiologists agreed were included in analyses

and those for which these radiologists disagreed were excluded from analyses. Our Institutional Review Board reviewed and approved the study protocol (protocol #142040).

Statistical analysis

Descriptive statistics are presented as mean and standard deviation (SD) or median and interquartile range [IQR], as appropriate for contin- uous data, or as proportions. We first used the Chi-square test to exam- ine univariate associations between patient characteristics and [1] identification of radiographically significant PNA and [2] administration of antibiotics. Multiple logistic regression models were then used to ex- amine adjusted associations between patient characteristics and perfor- mance of CXR for identification of PNA and between CXR interpretation and administration of antibiotics. Variables chosen for inclusion in the multivariable model were selected based on two factors: those with the most significant associations in the univariate analysis and the clin- ical findings considered to be most relevant in diagnosing pneumonia and thus most influential to a clinician’s decision to perform CXR. The number of variables entered in to these models was limited in order for there to be at least 10 outcomes (e.g., pneumonia on CXR) for each degree of freedom (d.f.) of the predictor variables in the model to avoid overfitting [12,13]. Because there were 44 patients with CXR- diagnosed pneumonia, we included the three variables with the lowest p-values in univariate analyses. Statistical analyses were performed using Stata 14.1 (StataCorp, College Station, TX). A Kappa coefficient was calculated to measure the inter-rater reliability of the two reading pediatric radiologists for all indeterminate chest radiographs. A random sampling of 10% of the initial 299 patient radiographs was provided to both radiologists for this purpose.

Results

A total of 527 charts were screened for inclusion. The flow chart (Fig. 1) displays the initial screening process. Forty-three patients (15%) had evidence of radiographic pneumonia on CXR and 51 (17.8%) were ulti- mately treated with antibiotics by the ED clinician for PNA. Agreement was found on 80% of the randomly sampled radiographs, resulting in a Kappa coefficient of 0.302. With expected agreement of 71.3% the ob- served agreement was fair and allowed us to reject the hypothesis that determination of radiographic pneumonia by each radiologist was random. CXRs of eleven additional patients who had ‘indeterminate’ in- terpretations of PNA by the pediatric radiologist at the time of the PED visit were excluded from analyses due to disagreement between the pe- diatric radiologists who adjudicated indeterminate interpretations for this study.

A total of 288 charts were included in the final analyses (Table 1), with median [IQR] age 7 [4, 10.5] years, 60% White, 49.5% African American, and 64.3% male. Sixty-two percent had public or no insurance while the remainder had private or military insurance. Fewer than 11% had hypoxia (defined as oxygen saturation on room air of b 92%), 21% had documented fever greater than or equal to 38 ?C and 5.5% had crackles documented on pulmonary examination. N 95% presented from an outpatient clinic or walk-in to the pediatric emergency depart- ment and 11.5% of all patients had been treated with an oral antibiotic within seven days of presentation for presumed pneumonia by a non- emergency room provider.

Radiographically significant PNA was present in 15%, and 17.8% of all patients had antibiotic initiated in the PED (Table 2). Five patients with radiographic pneumonia did not have antibiotics initiated and 13 pa- tients with no radiographic evidence of pneumonia were treated with antibiotics for either suspected atypical pneumonia or lobar pneumonia. One patient was found to have pneumothorax and did not require fur- ther clinical or surgical intervention as a result of this finding.

Univariate associations of patient characteristics with PNA on CXR

and initiation of antibiotic treatment in the PED are displayed in

Fig. 1. Patient Inclusion Process.

Table 3. Recent antibiotic use was associated with radiographic diagno- sis of pneumonia (p = 0.002), and continued antibiotic use (p = 0.002) and presence of crackles on physical examination (p = 0.03) was asso- ciated with continued use or initiation of antibiotic. Presence of hypoxia (oxygen saturations b 92% in triage), crackles on physical examination and recent use of antibiotics were included in the multivariable logistic regression model as they had the lowest p-values with PNA on CXR and subsequent management change in univariate models (p values 0.1, 0.5). In the multivariable logistic regression model adjusted for these three variables (Table 4), recent antibiotic use was associated with a positive CXR for pneumonia with an adjusted odds ratio (aOR) of 3.6 (95% CI 1.5, 9.0) and change in patient management by initiation of an- tibiotic aOR 3.2 (95% CI 1.4, 7.7). With respect to PNA on CXR and ad- ministration of antibiotics, hypoxia on presentation had aOR of 2.3 (95% CI 0.92, 5.8) and 1.5 (95% CI 0.59, 3.93) while crackles on examina- tion resulted in aORs of 2.1 (95% CI 0.64, 7.0) and 2.5 (95% CI 0.81, 7.7).

Table 1

Demographics and clinical Information in 288 children with acute asthma exacerbations in a pediatric emergency department?.

Age, years, median [interquartile range] 7 [4, 10.5] Race

White 115 (40)

Discussion

Performance of chest radiography in the evaluation of pediatric asth- ma exacerbations in the emergency department infrequently results in the identification of radiographically significant pneumonia and subse- quent clinical management change by administration of antibiotic by an emergency clinician. Our study suggests that after adjustment for other variables that might influence a clinician’s decision to obtain a CXR, sta- tistically significant associations with the presence of pneumonia or management change was found only when the patient had been treated with antibiotics for pneumonia within the previous seven days prior to emergency room presentation.

Common abnormalities seen on CXR in asthmatic patients are hy- perinflation, subsegmental atelectasis and peribronchial thickening. These findings do not typically correlate with the severity of an acute exacerbation or require clinical management changes. Pneumonia re- mains the most commonly diagnosed serious bacterial infection in chil- dren [14] and previous reports have observed that among asthmatic patients with radiographic abnormalities on CXR, pneumonia is the most commonly reported condition that results in a modification of

Table 2 Presence of Radiographic Pneumonia and antibiotic initiation in PED in 288 children with acute asthma exacerbations who had chest radiography performed?.

African-American 143 (49.5)

Male gender 185 (64)

Insurance Status

Public or none 178 (62)

Tricare 8 (2.7)

Hypoxia on arrival (b 92%) 32 (11)

Mode of arrival to PED

Walk-in/Clinic referral 275 (95)

Documented fever (N 38o C) 227 (79)

Crackles present on pulmonary exam 16 (5.5)

Antibiotic within 7 days of PED presentation 33 (11.5)

* Values are n (%) unless otherwise indicated.

PNA 43 (15)

Treatment with Antibiotic by PED Clinician 56 (19.4) Pneumonia on CXR

Antibiotic initiated 38 (13.3)

Antibiotic not initiated 5 (1.7)

No pneumonia on CXR

Antibiotic initiated 13 (4.5)

Antibiotic not initiated 232 (80.5)

PED, pediatric emergency department.

* Values are n (%) of cohort of 288 patients. Abbreviations: CXR, chest radiography; PNA, pneumonia on chest radiography as interpreted by pediatric radiologist.

Table 3

Univariate associations of patient characteristics with Radiographically Significant Pneu- monia on CXR and Treatment with Antibiotic in 288 children with acute asthma exacerba- tions in a pediatric emergency department?.

pneumonia and findings on chest radiography in Pediatric asthma pa- tients and subsequent typical management changes. The association found here suggests that 1) outpatient clinicians are confident in their

clinical diagnosis of pneumonia and 2) there may be an as-yet

Patient Variable	Pneumonia on CXR	Antibiotic Treatment
Age	0.99	1.0
Race	0.85	0.84
Gender	0.93	0.69
Insurance Status	0.46	0.63
Hypoxia (b92%)	0.1	0.5
PED Mode of Arrival	0.44	0.08
Fever N 38oC	0.5	0.23
Crackles	0.07	0.03
Antibiotic Use within 7 days of PED presentation	0.002	0.002

established overlapping window between clinical and radiographic di- agnosis of PNA in asthma patients which may be clinically useful if more definitively established in further study.

Conversely, however, patients presenting with asthma exacerbation

have acute airway changes secondary to bronchospasm, hyperresponsiveness, and mucous plugging [17]. As a result of airway inflammation, mucosal edema and mucous plugging, atelectasis may occur and cause opacification on CXR. This can result in a false positive diagnosis of pneumonia if radiography is obtained early during acute asthma treatment before atelectasis resolves. Bronchodilator and oxy-

Values are P values from chi-square test.

* CI = Confidence Interval.

usual asthma care. Radiographically significant pneumonia was present in 15% of our cohort, which is notably higher than previously reported [2,4,15,16]. This may be confounded by inclusion of patients in a tertiary PED where evaluation and referral of more severely ill patients is com- mon. Notwithstanding the higher prevalence of PNA on CXR, this find- ing was infrequently associated with a change in antibiotic management.

To our knowledge this is the first report of CXR findings documented exclusively from pediatric emergency department encounters for acute asthma exacerbations. In a study by Narayanan and colleagues, 405 pe- diatric charts for patients hospitalized with asthma exacerbation were examined retrospectively with one hundred and eighty (44%) contain- ing a CXR as part of their evaluation. The reported incidence of CXR- confirmed pneumonia treated with antibiotics was 3% and they report- ed three cases pneumothorax or pneumomediastinum (1.6%), none of which required additional Clinical interventions. Most CXRs in Narayanan’s study (74%) were performed in the emergency depart- ment/acute setting. They reported an incidence of fever at home or in the PED and triage saturations less than or equal to 92% were reported to be significantly associated with treatment-altering radiographic find- ings in pediatric inpatients with asthma exacerbation [15]. Previous studies report that b 7% of asthmatic patients are found to have an infil- trate on CXR that is consistent with clinically significant Bacterial pneumonia.

Remarkably, the 15% incidence of pneumonia in our study was twice that of previously reported studies. However, almost 5% of patients with a negative CXR were treated by PED clinicians for pneumonia, suggest- ing that a diagnosis and/or treatment of pneumonia still influenced by clinical assessment. This is also likely reflected in the results of our uni- variate analyses in regards to the presence of hypoxia on presentation and crackles on examination of this patient population. While both characteristics had p-values which approached but did not reach statis- tically significant values, persistence of either significant hypoxia or crackles on re-examination after sufficient bronchodilator therapy may motivate clinicians to obtain chest radiography as a result of their clinical gestalt. This is also the first study which examines the potential correlation between previous antibiotic use for clinically diagnosed

Table 4 Adjusted odds ratios for patient characteristics with pneumonia on chest radiography and antibiotic initiation for pneumonia in 288 children with acute asthma exacerbations who had chest radiography performed?.

Characteristic Pneumonia on CXR Antibiotic Initiation

	aOR*	95% CI		aOR	95% CI
Hypoxia	2.3	0.92, 5.8		1.5	0.59, 3.93
Crackles	2.1	0.64, 7.0		2.5	0.81, 7.7
Recent Antibiotic Use	3.6	1.5, 9.0		3.3	1.4, 7.7

Abbreviations: CI, confidence interval.

* Adjusted Odds Ratio.

gen treatments can also result in V/Q mismatch and relative hypoxia which may also prompt a clinician to obtain a CXR too soon. Our results suggest that even in the presence of fever (N 38C), hypoxia (b 92%)and crackles, CXR does not provide clinically significant information or man- agement changes and should be reserved for cases in which these find- ings persist despite bronchodilator treatment. However, it should be noted that the 95% confidence intervals of the adjusted odds ratios for the association of hypoxia (95% CI 0.92, 5.8) and of crackles (95% CI 0.64, 7.0) with pneumonia on CXR suggest that these are clinically meaningful associations and that our cohort did not have sufficient power to detect these associations at our level of statistical significance. While fever can also be clinical a manifestation of pneumonia, this find- ing should be interpreted with caution in pediatric patients with asth- ma. Approximately 80% of these exacerbations are triggered by viral respiratory infections, and fever is a frequent accompaniment of these infections [6,18]. Thus, the presence of fever alone does not warrant per- formance of CXR in pediatric patients with acute exacerbations.

Our study has limitations. First, the study used retrospective data that was not recorded for research purposes and that may have missing or incorrect data. The nature of retrospective studies has inherent risk for selection bias as patients are ultimately included based on their final diagnoses. Secondly, the ICD-9 codes used for screening charts for study inclusion have not, to our knowledge, been previously validat- ed for study purposes thus may have resulted in exclusion of appropri- ate study patients. Thirdly, interpretation of CXRs for PNA is subjective, and experienced pediatric radiologists may not consistently agree on this finding. Indeed, the Kappa coefficient between our two adjudicating radiologists was 0.30, though these individuals were adjudicating the 4% of cases in which the initial radiologist interpretation was already in- determinate. Lastly, our multivariable logistic models were limited by the number of outcomes (pneumonia on CXR) and could accommodate only three covariates to avoid overfitting. While these covariates the greatest univariate associations with our outcome of interest, other clin- ical factors may influence the clinician’s decision for a CXR in this population.

Previous studies examining the relevance of CXR in pediatric pa- tients with acute asthma exacerbations have included hospitalized and emergency department patients and have produced similar results thus supporting recommendations that chest radiography not be part of an initial diagnostic work up for AAE. Our study uniquely examines po- tential clinical and demographic factors exclusively in emergency de- partment setting, making it more practical for clinicians in the emergency department where high-volume, high-turnover patient care requires streamlined diagnostic decision making. The results of our study have potential reduce unnecessary ancillary testing, lower overall healthcare costs, and unnecessary radiation exposure to pediat- ric patients with uncomplicated acute asthma exacerbation.

Conclusion

CXR infrequently identifies pneumonia or results in change in stan- dard management of children with acute asthma exacerbations. This

imaging may more appropriately be deferred for a period of time to de- termine if the patient improves after asthma-specific treatment. Recent antibiotic treatment within several days prior to PED presentation for AAE may be predictive of positive CXR findings and subsequent contin- uation of treatment for PNA but a larger study is needed to determine the potential value in the presence of hypoxia and crackles on presentation.

Funding Source

None.

References

1. Akinbami LJ, Moorman JE, Liu X. Asthma prevalence, Health care use, and mortality: United States, 2005-2009. Natl Health Stat Report 2011(32):1-14.
2. Mathews B, Shah S, Cleveland RH, Lee EY, Bachur RG, Neuman MI. Clinical predictors of pneumonia among children with wheezing. Pediatrics 2009;124(1):e29-6.
3. Swischuk LE. Asthma attack: is a chest X-ray necessary? Pediatr Emerg Care 2005; 21(7):468-70.
4. Gershel JC, Goldman HS, Stein RE, Shelov SP, Ziprkowski M. The usefulness of chest radiographs in first asthma attacks. J Med]->N Engl J Med 1983;309(6):336-9.
5. Stanley RM, Teach SJ, Mann NC, Alpern ER, Gerardi MJ, Mahajan PV, et al. Variation in ancillary testing among pediatric asthma patients seen in emergency departments. Acad Emerg Med 2007;14(6):532-8.
6. Heymann PW, Carper HT, Murphy DD, Platts-Mills TA, Patrie J, McLaughlin AP, et al. viral infections in relation to age, atopy, and season of admission among children hospitalized for wheezing. J Allergy Clin Immunol 2004;114(2):239-47.
7. Khetsuriani N, Kazerouni NN, Erdman DD, Lu X, Redd SC, Anderson LJ, et al. Preva- lence of viral Respiratory tract infections in children with asthma. Journal of Allergy and Clinical Immunology 2007;119(2):314-21.
8. Fitzwater JW, Silva NN, Knight CG, Malvezzi L, Ramos-Irizarry C, Burnweit CA. Man- agement of spontaneous pneumomediastinum in children. J Pediatr Surg 2015; 50(6):983-6.
9. Abbas PI, Akinkuotu AC, Peterson ML, Mazziotti MV. Spontaneous pneumomediastinum in the pediatric patient. J Surg]->Am J Surg 2015;210(6): 1031-5 [discussion 5-6].
10. Dotson K, Timm N, Gittelman M Is. Spontaneous pneumothorax really a pediatric problem? A national perspective. Pediatr Emerg Care 2012;28(4):340-4.
11. Taussig LM, Wright AL, Holberg CJ, Halonen M, Morgan WJ, Martinez FD. Tucson Children’s respiratory study: 1980 to present. J Allergy Clin Immunol 2003;111(4): 661-75 [quiz 76].
12. Harrell FE, Jr., Lee KL, Mark DB. Multivariable prognostic models: issues in develop- ing models, evaluating assumptions and adequacy, and measuring and reducing er- rors. Stat Med 1996;15(4):361-87.
13. Steyerberg E. Clinical prediction models: A practical approach to development, val- idation, and updating. New York, NY: Springer; 2009.
14. Florin TA, French B, Zorc JJ, Alpern ER, Shah SS. Variation in emergency department diagnostic testing and disposition outcomes in pneumonia. Pediatrics 2013;132(2): 237-44.
15. Narayanan S, Magruder T, Walley SC, Powers T, Wall TC. Relevance of chest radiog- raphy in pediatric inpatients with asthma. J Asthma 2014;51(7):751-5.
16. Brooks LJ, Cloutier MM, Afshani E. Significance of roentgenographic abnormalities in children hospitalized for asthma. Chest 1982;82(3):315-8.
17. Carroll CL, Sala KA. Pediatric Status asthmaticus. Crit Care Clin 2013;29(2):153-66.
18. Khetsuriani N, Kazerouni NN, Erdman DD, Lu X, Redd SC, Anderson LJ, et al. Preva- lence of viral respiratory tract infections in children with asthma. J Allergy Clin Immunol 2007;119(2):314-21.