Prediction of Pneumonia in a Pediatric EmergencyDepartment
WHAT’S KNOWN ON THIS SUBJECT: Use of chest radiography inthe evaluation of children with possible pneumonia varies widely.Although studies have identified certain historical features andphysical examination findings associated with pneumonia, nonehave specifically addressed the criteria for obtaining a chestradiograph.
WHAT THIS STUDY ADDS: Clinical data can stratify children forpneumonia risk. Children with hypoxia and focal lung findings arehigh risk whereas those without hypoxia, fever, and ausculatoryfindings are low risk. For low-risk patients, clinical follow-upshould be considered over obtaining a radiograph.
abstractOBJECTIVE: To study the association between historical and physicalexamination findings and radiographic pneumonia in children whopresent with suspicion for pneumonia in the emergency department,and to develop a clinical decision rule for the use of chest radiography.METHODS: We conducted a prospective cohort study in an urban pediat-ric emergency department of patients younger than 21 who had a chestradiographperformed for suspicionof pneumonia (n�2574). Pneumoniawas categorized into 2 groups on the basis of an attending radiologistinterpretation of the chest radiograph: radiographic pneumonia (includesdefinite and equivocal cases of pneumonia) and definite pneumonia. Weestimated amultivariate logistic regressionmodelwith pneumonia statusas the dependent variable and the historical and physical examinationfindings as the independent variables. We also performed a recursivepartitioning analysis.
RESULTS: Sixteen percent of patients had radiographic pneumonia. His-tory of chest pain, focal rales, duration of fever, and oximetry levels attriage were significant predictors of pneumonia. The presence of tachy-pnea, retractions, and gruntingwere not associatedwith pneumonia. Hyp-oxia (oxygen saturation�92%)was the strongest predictor of pneumonia(odds ratio: 3.6 [95% confidence interval (CI): 2.0–6.8]). Recursive parti-tioning analysis revealed that among subjects with O2 saturation�92%,no history of fever, no focal decreased breath sounds, and no focal rales,the rate of radiographic pneumonia was 7.6% (95% CI: 5.3–10.0) and defi-nite pneumonia was 2.9% (95% CI: 1.4–4.4).
CONCLUSION: Historical and physical examination findings can be usedto risk stratify children for risk of radiographic pneumonia. Pediatrics2011;128:246–253
AUTHORS: Mark I. Neuman, MD, MPH,a Michael C.Monuteaux, ScD,a,b Kevin J. Scully, BS,a and Richard G.Bachur, MDa
aDivision of Emergency Medicine and bClinical ResearchProgram, Children’s Hospital Boston, Boston, Massachusetts
FINANCIAL DISCLOSURE: The authors have indicated they haveno financial relationships relevant to this article to disclose.
246 NEUMAN et alby guest on July 14, 2016Downloaded from
The use of chest radiography in theevaluation of children with possiblepneumonia varies widely. The reasonsfor this variation include the lack of agold standard for the diagnosis ofpneumonia, difficulties in appreciatingsubtle signs of pneumonia (particu-larly in young children and infants),and the differential availability of radi-ology across practice settings. In addi-tion, variability exists in the interpreta-tion of chest radiographs (CXRs) forthe diagnosis of pneumonia, and evenunder ideal circumstances it is diffi-cult to distinguish viral from bacterialpneumonia solely on the basis of theCXR alone.
Studies that have sought to developclinical decision rules for the evalua-tion of children with suspected pneu-monia have been limited by their retro-spective nature or small samplesize.1–6 Although in these studies cer-tain historical features and physicalexamination findings associated withpneumonia have been identified, in fewhas the criteria for obtaining a CXRbeen specifically addressed.
We have conducted the largest pro-spective evaluation of childrenwho un-derwent radiography for the suspicionof pneumonia in the emergency de-partment (ED) setting to better identifypatients at both low and high risk ofradiographic pneumonia. We sought toassess the relation between historicaland physical examination findings andradiographic pneumonia, and to de-velop a clinical decision rule to guidephysicians in the use of radiographyfor children at risk of pneumonia.
METHODS
Study Design
We conducted a prospective cohortstudy in an urban pediatric ED with�56 000 visits annually. Childrenyounger than 21 who underwent a CXRfor the evaluation of possible pneumo-nia were included in the study. Pa-
tients were excluded from the study ifthey had a CXR for an indication otherthan suspicion of pneumonia or if theyhad a significant previous medical his-tory that would predispose a patienttoward pneumonia, such as sickle celldisease, cardiac disease, immunodefi-ciency, or severe neurologic disorder.The study took place between Novem-ber 2006 and May 2009.
All physicians who worked in the EDwere asked to participate in the studyand were informed about the study de-tails before involvement. Physicianscompleted a brief questionnaire abouttheir patient’s presentation after re-questing a CXR for suspicion of pneu-monia but before viewing the radio-graph or obtaining a reading fromradiology. After completion, question-naires were placed in secure lock-boxes located throughout the ED. Allphysicians who participated in thestudy were board-certified pediatricemergency medicine physicians orgeneral pediatricians. Questionnairescompleted by residents required thereal-time review and signature of theattending physician to verify the data.
Definitions
The identification of pneumonia wasbased on the final attending pediatricradiologist’s report in the electronicmedical chart. A patient was consid-ered to have radiographic pneumoniaif the CXR had definite findings of pneu-monia, and also included radiographswith equivocal findings of pneumonia.Within the subset of patients with ra-diographic pneumonia, we defined agroup of patients with definite pneu-monia, which included children withCXR reports with descriptors such as“consolidation,” “infiltrate” or “pneu-monia.” Radiographic findings ofequivocal pneumonia included thosewith descriptors such as “atelectasisversus infiltrate,” “atelectasis versuspneumonia,” or “likely atelectasis but
cannot exclude (or rule out) pneumo-nia.” If the CXR reading included termi-nology such as “normal chest,” “nor-mal radiograph,” “clear lungs,” “noacute pulmonary findings,” “atelecta-sis,” or “peribronchial cuffing,” it wasconsidered negative for pneumonia.
Data Collection
There were 2 primary mechanisms fordata collection in our study: informa-tion from prospectively collected ques-tionnaires and medical chart review.Questionnaires asked physicians to re-mark on such factors as the patient’sappearance, level of respiratory dis-tress, symptoms (fever, cough, wheez-ing, chest pain, difficulty breathing),and the reason(s) for obtaining theCXR (height and duration of fever, re-spiratory distress, cough, hypoxia, in-creasedwhite blood cell count). Partic-ular attention was paid to thepresence and location of physical ex-amination findings, such as retrac-tions, grunting, focal decreasedbreath sounds, wheezing, and rales.The questionnaires were designed inthe fixed-choice format to prevent freetext responses (Fig 1).
The electronic medical charts were re-viewed by the study investigators toobtain basic demographic information(eg, age, gender), vital signs (eg, tem-perature, oxygen saturation, respira-tory rate), treatment in the ED, dispo-sition, radiograph results, and finaldiagnosis. Tachypnea was defined byage-specific thresholds as measuredby the respiratory rate at triage.6,7 Themedical charts were also reviewed toassess for comorbid conditions thatpredispose toward pneumonia, includ-ing asthma, reactive airway disease,and bronchiolitis.
To evaluate enrollment bias, we au-dited daily radiography logs for thefirst 3 days of each month during thefirst year of the study period. We esti-mated the enrollment rate by dividing
ARTICLES
PEDIATRICS Volume 128, Number 2, August 2011 247by guest on July 14, 2016Downloaded from
the total numbers of study subjects bythe total number of eligible patientswho had a CXR performed to evaluatefor pneumonia.8 The rates of pneumo-nia were then calculated for enrolledand not-enrolled children.
Data Analysis and StatisticalMethods
All data were entered into a secureelectronic database with data valida-tion rules. Descriptive and multivari-ate data analysis were performed withStata 10.1 (Stata Corp, College Station,TX). First, we assessed the relationbetween historical features and physi-cal examination findings and radio-graphic pneumonia. We estimated amultivariate logistic regression model
with pneumonia status (radiographicpneumonia versus no pneumonia) asthe dependent variable, and the inde-pendent variables were the historical(difficulty breathing, chest pain, dis-tress, cough, fever) and physical exam-ination findings (tachypnea defined byage-specific respiratory rate thresh-olds measured at triage, retractions,grunting, focal decreased breathsounds, rales, focal rales, focal wheez-ing, triage temperature, and room airoximetry value). Temperature at triagewas dichotomized (�38°C or �38°),whereas history of fever, history ofcough (none, up to 72 hours, or �72hours), and oxygen saturation at tri-age (97%–100%, 93%–96%, or �92%)
were modeled as categorical vari-ables. Hypoxia was defined as oxygensaturation�92%. We also estimated amodel with definite pneumonia as thedependent variable and the same his-torical and physical examination find-ings as the independent variables. Theodds ratios (OR) with a 95% confidenceinterval (CI) was calculated for all vari-ables. All statistical tests were 2-tailed.
Because data on oxygen saturation attriage were missing on 199 subjects(5%), we repeated the multivariateanalysis while assigning these pa-tients a value for this variable that wasoutside the observed range (ie, a valueequal to the lowest observed value mi-nus 1) and including a binary variable
FIGURE 1Data collection form.
248 NEUMAN et alby guest on July 14, 2016Downloaded from
that indicated the presence or ab-sence of an observed oximetry value.Although sacrificing precision by intro-ducing an additional covariate, thismethod allowed us to generate modelestimates using these 199 patients.
We then developed a clinical decisiontree to risk-stratify patients being con-sidered for pneumonia and therebyguide clinicians around the use of ra-diography. We performed a recursivepartitioning analysis using CART 5 soft-ware (Salford Systems, Stanford, CA).Variables that were significant or nearsignificant (P� .20) in themultivariatelogistic model described above weredesignated as candidate variables forinclusion in the recursive partitioninganalysis. We applied a cost ratio of 5:1for falsely categorizing a patient as nothaving pneumonia (ie, false-negative)compared with misclassifying a pa-tient with radiographic pneumonia (ie,false-positive). The “optimal” tree asdetermined by the default CART algo-rithm was presented. We then per-formed multivariate analyses and re-cursive partitioning analysis on thesubset of children younger than 5. Theinstitutional review board approvedthis study. Data collection was compli-ant with the Health Insurance Portabil-ity and Accountability Act of 1996.
RESULTS
Subjects
The demographic and clinical char-acteristics of the sample are shownin Table 1. The majority of the samplewas younger than 5, with a median ageof 2.3 years. Subjects were slightlymore likely to be male (54%). At triage,20%, 5%, and 37% of subjects pre-sented with age-adjusted tachypnea,hypoxia, and pyrexia, respectively.Seventy-five percent and 91% of pa-tients reported a history of fever andcough, respectively. Wheezing was ob-served in 27% of patients. The propor-tion of patients with radiographic
pneumonia was 16%, whereas 8%were classified as having definitepneumonia. Radiographic findingsamong children with definite pneumo-nia included lobar consolidation(73%), multilobar consolidation (21%),and consolidation with pleural effu-sion (6%). Overall, 22% of patientswere hospitalized.
Enrollment
Overall, 51% of eligible patients wereenrolled on the basis of review onradiology logs for the first 3 days of
each month during the first 12months of the study period. On thebasis of our sampling, patients eligi-ble but not enrolled did not differfrom enrolled patients with respectto age (median age of 2.3 years inboth groups) or the rate of definitepneumonia (6.2% versus 6.0%, re-spectively, P � .5).
Multivariate Regression
Our multivariate model of the clinicalpredictors of radiographic and defi-nite pneumonia is displayed in Table 2.
TABLE 1 Demographics of Study Population (n� 2574)
Characteristic
Age, median (IQR), y 2.3 (0.9–5.2)Age, N (%), y
�2 1189 (46.2)2–4.9 712 (27.7)5–9.9 401 (15.6)10–21.9 272 (10.6)Male, N (%) 1381 (53.7)Triage temperature, N (%), °C
97–100 1663 (70.0)93–96 593 (25.0)�92 119 (5.0)Duration of fever, N (%)None 651 (25.3)�72 h 1497 (58.2)�72 h 426 (16.6)Duration of cough, N (%)a
None 224 (8.8)�72 h 1346 (52.9)�72 h 974 (38.3)Wheezing present upon examination in ED, N (%) 704 (27.4)Patients with WBC done, N (%) 749 (29.1)WBC count, mean (SD) 13.1 (7.1)Radiographic pneumonia, N (%)Definite pneumonia 199 (7.7)Radiographic pneumonia 422 (16.4)% Admitted to hospital, N (%) 576 (22.4)Definite pneumonia 67 (34.2)Radiographic pneumonia 133 (31.7)
IQR indicates interquartile range; WBC, white blood cell; RR, respiratory rate.a Duration of cough was missing in 30 patients.b Triage oxygen saturation was missing in 199 patients.
ARTICLES
PEDIATRICS Volume 128, Number 2, August 2011 249by guest on July 14, 2016Downloaded from
As shown, children with a history ofchest pain and focal rales on examina-tion were significantly more likely tobe diagnosed with definite pneumoniarelative to children without these clin-ical features. Conversely, children whopresented with wheezing on examina-tion were significantly less likely to bediagnosed with definite pneumonia.We also found that the duration of fe-ver was positively correlated with theodds of a definite pneumonia diagno-sis. In addition, children with fever�38°C or hypoxia at triage were at asignificantly increased risk for definitepneumonia. When we repeated thismodel while including the 199 subjectswith missing information on oxygensaturation at triage, the pattern of re-sults did not change.
The results of amultivariatemodel of theclinical predictors of radiographic pneu-monia were similar to the model that
predicted definite pneumonia, with his-tory of chest pain, focal rales, durationoffever, and oxygen saturation levels at tri-age emerging as significant predictors.Subjects with wheezing on examinationwere at lower risk of radiographic pneu-monia. However, in contrast to themodelthat predicted definite pneumonia, tri-age temperature was not a significantpredictor of radiographic pneumonia(P� .08). When we repeated this modelwhile including the 199 subjects withmissing information on oxygen satura-tion at triage, the patterns of results didnot change, with the exception of focaldecreased breath sounds, which was asignificant predictor of radiographicpneumonia (OR: 1.39 [95%CI: 1.02–1.90]),P� .04).
Development of Decision Tree
On the basis of the results of the multi-variate logistic regression model, the
follow factors were selected for inclu-sion in the recursive partitioning analy-sis predicting radiographic pneumonia:wheezing on examination, chest pain, fo-cal decreased breath sounds, focalrales, triage temperature, triage oxygensaturation, and history of fever. Asshown in Fig 2, oxygen saturation ini-tially divided the sample into “highrisk” (O2 saturation, �92%; rate ofpneumonia, 37%) and “intermediaterisk” (O2 saturation, �92%; rate ofpneumonia, 15%) nodes. Additionalsplits of the “intermediate risk” nodeby history of fever, focal decreasedbreath sounds, and focal rales identi-fied a lower risk group (n� 503). Thatis, among subjects with O2 saturation�92%, no history of fever, no focal de-creased breath sounds, and no focalrales, the rate of radiographic pneu-monia was 7.6% (95% CI: 5.3–10.0). Therate of definite pneumonia among thissame subsample was 2.9% (95% CI:1.4–4.4).
Patients Younger Than 5
In a multivariate model that pre-dicted definite pneumonia in childrenyounger than 5 (n� 1901), the patternof results was the same, with the ex-ception of wheezing on examination,which was no longer significant (OR:0.64 [95% CI: 0.38–1.10] P � .10). In amodel that predicted radiographicpneumonia, both chest pain (OR: 1.86[95% CI: 0.96–3.60] P � .06) andwheezing on examination (OR: 0.78[95% CI: 0.56–1.10] P � .16) were nolonger statistically significant. The re-maining pattern of results was thesame. In a recursive partitioning anal-ysis, oxygen saturation remained thesingle best predictor of radiographicpneumonia in children younger than 5.Thirty-nine percent of children with O2saturation �92% had pneumonia,whereas 15% of children with O2 satu-ration �92% had pneumonia. Al-though the intent was to identify a low-risk group, we were unable to further
TABLE 2 Triage and Clinical Predictors of Pneumonia in Multivariate Analyses
All of the predictors listed were included simultaneously in both Model 1 and Model 2.a 222 study subjects weremissing values for either oxygen saturation at triage or duration of cough andwere excluded frommultivariate analysis.b Determined by parental report.
250 NEUMAN et alby guest on July 14, 2016Downloaded from
characterize a low risk populationamong children younger than 5.
DISCUSSION
Historical and physical examinationfindings can be used to stratifychildren for risk of radiographicpneumonia. Unfortunately, the testcharacteristics of individual physicalexamination findings, such as focalrales, lack adequate sensitivity andspecificity to confirm or exclude thediagnosis of pneumonia. Certaincharacteristics such as hypoxia, lackof wheeze, and focal rales place chil-dren at increased risk of radio-graphic pneumonia, whereas therate of pneumonia is lower in the ab-
sence of hypoxia and fever, and with-out focal ausculatory findings.
The development of a clinical practiceguideline for the use of chest radiogra-phy for the diagnosis of pneumonia inchildren may improve the quality ofcare and reduce CXR use. Identificationof a lower-risk population may help toreduce unnecessary testing and radia-tion exposure, whereas identificationof a high risk group will help ensurethat radiography is performed toconfirm a suspected diagnosis ofpneumonia. Many previous studiesthat have sought to predict pneumo-nia in children have been limited bytheir small sample size1,9,10 or select
patient population,3,4,6,7 or have beenconducted in resource poor set-tings,11,12 where the rate of pneumo-nia is considerably higher than in in-dustrialized nations.
No single or combination of physicalexamination findings will have per-fect sensitivity for the identificationof pneumonia in children. Occultpneumonia or radiographic pneumo-nia in a child without respiratory dis-tress or ausculatory findings on ex-amination is well described in up to5% to 10% of children for whom aCXR is obtained.6,13,14 These observa-tions highlight the challenges in thedevelopment of a highly sensitive
Eligible patients(n=2574)
PNA (n=422, 16.4%)
O2 sat < 92%(n=119)
PNA (n=44, 37.0%)
O2 sat > 92%(n=2455)
PNA (n=378, 15.4%)
No wheeze on exam(n=61)
PNA (n=31, 50.8%)
Wheeze on exam(n=58)
PNA (n=13, 22.4%)
Focal rales(n=20)
PNA (n=14, 70.0%)
No focal rales(n=41)
PNA (n=17, 41.5%)
History of fever(n=1840)
PNA (n=321, 17.4%)
No history of fever (n=615)
PNA (n=57, 9.3%)
Chest pain(n=162)
PNA (n=46, 28.4%)
No chest pain(n=1678)
PNA (n=275, 16.4%)
Focal dec BS(n=65)
PNA (n=12, 18.5%)
No focal dec BS(n=550)
PNA (n=45, 8.2%)
Focal rales(n=47)
PNA (n=7, 14.9%)
No focal rales(n=503)
PNA (n=38, 7.6%)
FIGURE 2Stratification of patients on the basis of risk of radiographic pneumonia using recursive partitioning analysis (n� 2574). We included variables that weresignificant at the P� .2 level in multivariate logistic regression model. We assigned a cost of 5:1 for falsely categorizing a patient as not having pneumoniacompared with misclassifying a patient with radiographic pneumonia. dec BS indicates decreased breath sounds.
ARTICLES
PEDIATRICS Volume 128, Number 2, August 2011 251by guest on July 14, 2016Downloaded from
clinical decision tool to help clini-cians manage children with sus-pected pneumonia.
In our study we expand and refine thefindings of other studies in whichpredictors of pneumonia in childrenhave been investigated. The majorstrength of our study lies in its pro-spective data collection and largesample size. Lynch et al2 found thefollowing findings to be associatedwith a focal infiltrate on CXR: historyof fever; tachypnea; retractions;grunting; rales; and decreasedbreath sounds. Oxygen saturationwas not studied because it was in-consistently recorded in their study,and 36% of their study populationhad radiographic pneumonia. Thehigh rate of pneumonia may be be-cause of their exclusion of young in-fants and children with asthma, inwhom the rate of radiographic pneu-monia is lower. More recently, Bilkiset al11 validated the decision rule de-scribed by Lynch et al2 that looked atthe combination of 4 findings (fever,localized rales, decreased breathsounds, and tachypnea) and pro-cured another decision rule. How-ever, in that investigation, 69% ofstudy subjects had radiographicpneumonia. Our study differs fromthese studies in that all children whohad a CXR performed for the evalua-tion of pneumonia were eligible forinclusion, which is reflected in ourlower rates of pneumonia (definitepneumonia, 8%; radiographic pneu-monia, 16%). We believe that our rateof radiographic pneumonia is withinthe typical range of other studies inchildren,1,5,10,15 which may make ourfindings more generalizable to a pop-ulation of children for whom a CXR isobtained to evaluate for pneumoniain an ED setting. In one such study,Mahabee-Gittens et al5 found that ox-ygen saturation�96%, nasal flaring,and age �12 months were associ-
ated with radiographic infiltratesamong children 2 to 59 months eval-uated in an ED setting.
There is wide variability in the man-agement of children suspected tohave pneumonia, which is under-standable given the wide variabilityin clinical findings in children withpneumonia, the lack of gold standardto establish the diagnosis, and diffi-culties in distinguishing patientswith viral and bacterial pneumonia.Rothrock et al9 evaluated Canadiantask force published guidelines fordiagnosing pneumonia, which con-cluded that the absence of each ofthe 4 signs (ie, respiratory distress,tachypnea, rales, and decreasedbreath sounds) accurately excludesthe diagnosis of pneumonia.16 Ro-throck et al9 noted that application ofthese criteria had a sensitivity of45% and specificity of 66% for the di-agnosis of pneumonia in an ED popu-lation of children. Our group hasdemonstrated that the use of tachy-pnea alone does not distinguish chil-dren with and without radiographicpneumonia when applied to a US-based ED setting, yet this is the majorscreening tool used by the WorldHealth Organization in resource poorsettings.8 As a result of the variabilityin clinical and radiographic findingsobserved in childhood pneumonia,many clinicians will treat patientsbased solely on clinical findings.17
For example, a child with history offever and focal rales on examinationwill likely be treated with antimicro-bial agents for suspected pneumo-nia, particularly in the outpatientsetting. However, our data revealthat only 25% of children with thiscombination of findings have a radio-graphic abnormality (radiographicpneumonia); 13% have pneumoniawhen using a stricter definition (def-inite pneumonia). Use of clinical cri-teria alone may be justified by the
benefit of avoiding potentially harm-ful radiation but should be weighedagainst the burden of inaccurate di-agnoses and unnecessary antibioticuse.
There are several limitations to ourstudy. Approximately half of the eligi-ble patients who underwent a CXR inthe ED were enrolled in the study.However, patients not enrolled didnot differ from those included in thestudy with respect to age and thepresence of radiographic pneumo-nia. Although it is unlikely that thereis enrollment bias in our sample, wewere unable to verify this by lookingat other objective parameters (eg,oxygen saturation, ausculatory find-ings). In addition, enrollment was fa-cilitated by the availability of re-search staff in the ED and was notrelated in a systematic way to pa-tients’ or physicians’ characteris-tics. Thus, any selection bias attrib-utable to our enrollment rate isunlikely to compromise the validityof our findings. The study was con-ducted in a single ED of a tertiarycare children’s hospital, which maylimit the generalizability to otherpractice settings. The entry criteriafor our study required the clini-cal suspicion of pneumonia thatprompted the decision to obtain a ra-diograph. We did not study allpatients presenting with cough orfever, which may limit the generaliz-ability of our findings. We enrolledchildren younger than 21, yet the de-cision to obtain a radiograph on thesubset of children younger than 5may be most challenging. Consider-ation of the wide age range of en-rolled children should be consideredwhen interpreting our findings. Weare also unable to evaluate the reli-ability of specific physical examina-tion findings because patients wereonly examined by the treating physi-cian and did not undergo a second
252 NEUMAN et alby guest on July 14, 2016Downloaded from
examination for the purpose of thisstudy. In addition, we were not ableto evaluate children in whom therewas suspicion of pneumonia but a ra-diograph was not obtained. However,recent data from our institution indi-cates that 83% of patients dis-charged from the ED with a diagnosisof pneumonia had a CXR obtained.Lastly, radiologists were not blindedto the clinical information obtained
by the physician whom assessed thelikelihood of pneumonia, which couldfurther bias our results.
CONCLUSION
Historical features and physical ex-amination findings can be used tostratify children for risk of radio-graphic pneumonia. Children withhypoxia and focal lung findings are athigh risk of radiographic pneumo-
nia, whereas the rate of pneumoniais low among children without hyp-oxia, fever, and localized ausculatoryfindings. For the low-risk patient,careful clinical follow-up should beconsidered as an alternative to ob-taining a radiograph. Validation ofthese findings requires multicenterresearch and should include the as-sessment of the reliability of physi-cal examination findings.
REFERENCES
1. Leventhal JM. Clinical predictors of pneu-monia as a guide to ordering chest roent-genograms. Clin Pediatr (Phila). 1982;21(12):730–734
2. Lynch T, Platt R, Gouin S, Larson C, Pat-enaude Y. Can we predict which childrenwith clinically suspected pneumonia willhave the presence of focal infiltrates onchest radiographs? Pediatrics. 2004;113(3pt 1). Available at: www.pediatrics.org/cgi/content/full/113/3/e186
3. Mahabee-Gittens EM, Bachman DT, ShapiroED, Dowd MD. Chest radiographs in the pe-diatric emergency department for children� or � 18 months of age with wheezing.Clin Pediatr (Phila). 1999;38(7):395–399
4. Mahabee-Gittens EM, Dowd MD, Beck JA,Smith SZ. Clinical factors associated withfocal infiltrates in wheezing infants and tod-dlers. Clin Pediatr (Phila). 2000;39(7):387–393
5. Mahabee-Gittens EM, Grupp-Phelan J, BrodyAS, et al. Identifying children with pneumo-nia in the emergency department. Clin Pe-diatr (Phila). 2005;44(5):427–435
6. Murphy CG, van de Pol AC, Harper MB,Bachur RG. Clinical predictors of occult
pneumonia in the febrile child. Acad EmergMed. 2007;14(3):243–249
7. MathewsB,ShahS,ClevelandRH,LeeEY,BachurRG, Neuman MI. Clinical predictors of pneumo-nia among children with wheezing. Pediatrics.2009;124(1). Available at: www.pediatrics.org/cgi/content/full/124/1/e29
8. Shah S, Bachur R, Kim D, NeumanMI. Lack ofpredictive value of tachypnea in the diagno-sis of pneumonia in children. Pediatr InfectDis J. 2010;29(5):406–409
9. Rothrock SG, Green SM, Fanelli JM, CruzenE, Costanzo KA, Pagane J. Do publishedguidelines predict pneumonia in childrenpresenting to an urban ED? Pediatr EmergCare. 2001;17(4):240–243
10. Zukin DD, Hoffman JR, Cleveland RH, Kush-ner DC, Herman TE. Correlation of pulmo-nary signs and symptoms with chest radio-graphs in the pediatric age group. AnnEmerg Med. 1986;15(7):792–796
11. Bilkis MD, Gorgal N, Carbone M, et al. Valida-tion and development of a clinical predic-tion rule in clinically suspected community-acquired pneumonia. Pediatr Emerg Care.2010;26(6):399–405
12. Oliveira LL, Silva SA, Ribeiro LH, de OliveiraRM, Coelho CJ, S Andrade AL. Computer-aided diagnosis in chest radiography fordetection of childhood pneumonia. Int JMed Inform. 2008;77(8):555–564
13. Bachur R, Perry H, Harper MB. Occult pneu-monias: empiric chest radiographs in fe-brile children with leukocytosis. Ann EmergMed. 1999;33(2):166–173
14. Shah S, Mathews B, Neuman MI, Bachur R.Detection of occult pneumonia in a pediat-ric emergency department. Pediatr EmergCare. 2010;26(9):615–621
15. Taylor JA, Del Beccaro M, Done S, Winters W.Establishing clinically relevant standardsfor tachypnea in febrile children youngerthan 2 years. Arch Pediatr Adolesc Med.1995;149(3):283–287
16. Jadavji T, Law B, Lebel MH, Kennedy WA, GoldR, Wang EE. A practical guide for the diagno-sis and treatment of pediatric pneumonia.CMAJ. 1997;156(5):S703–S711
17. Metlay JP, Kapoor WN, Fine MJ. Does thispatient have community-acquired pneumo-nia? Diagnosing pneumonia by history andphysical examination. JAMA. 1997;278(17):1440–1445
ARTICLES
PEDIATRICS Volume 128, Number 2, August 2011 253by guest on July 14, 2016Downloaded from
