Are we missing the mark? Fever, respiratory symptoms, chest radiographs, and acute chest syndrome in sickle cell disease

Abstract

Acute chest syndrome (ACS) is defined as a new pulmonary infiltrate detected by chest radiography (CXR) associated with fever, respiratory symptoms, or chest pain 1, 2. It is the second most common cause of hospital admissions in SCD, and can lead to significant morbidity and mortality 2. Evidence in the literature has established that clinical assessment alone is not a sensitive predictor of ACS 1, 2. Although these studies were performed nearly 20 years ago, they advocate for empiric CXR utilization all febrile children with SCD 1, 2. Since then, the incidence of ACS in febrile children has decreased from 27 to 17.4% (P < 0.001) 3, likely due in part to routine penicillin prophylaxis, expanded pneumococcal immunization coverage, and hydroxyurea use. National Institutes of Health (NIH) guidelines for febrile SCD patients were updated in 2014 recommending that patients with fever associated with shortness of breath, tachypnea, cough, and/or rales should undergo CXR to evaluate for ACS 4. No systemic review of the literature was done regarding these SCD/fever guidelines, which utilized panel consensus expert opinion. Since there are discrepancies between the current NIH guidelines, many institutional clinical pathway recommendations and evidence from the literature, this study was undertaken to help identify febrile children with SCD requiring a CXR as part of their assessment. 2013 emergency department (ED) data were obtained by electronic medical record using ICD-9 codes linked to fever and SCD in children 2 months–18 years through a retrospective chart review at two urban tertiary-care campuses of Children's Healthcare of Atlanta (CHOA). Demographics, past medical history, vital signs, review of systems (ROS), physical exam (PE) findings, CXR radiology interpretation, and ACS diagnosis were obtained by chart review; see “Supporting Information Methods” for details. A total of 356 patients with 609 ED encounters were evaluated. (see Supporting Information Fig. 1 for flowchart). Patient demographics are summarized in Supporting Information Table I. The mean age was 5.7 ± 5 years. Half were female and the majority of patients (67%) had Hb-SS. Patient demographics, clinical characteristics and risk factors for ordering of CXR are summarized in Supporting Information Table II. 379 CXRs were ordered from the 609 encounters (62%); see Supporting Information Methods/Results. Supporting Information Table II summarizes demographics, clinical characteristics and risk factors for patients with ACS (n = 66) compared to those with negative CXRs. Of encounters evaluated by CXR, the prevalence of ACS was 17.2% in our cohort. Risk factors for ACS as determined by univariate analysis include the presence of tachypnea, history of ACS, and abnormal ROS (presence of cough, wheeze, or chest pain). An abnormal lung exam was also more frequently found in children diagnosed with ACS. However, 62% of patients with ACS had a normal lung exam. Patients with ACS were also more frequently admitted to the hospital. Of all admitted patients, those with ACS more frequently utilized oxygen and BiPAP. However, there was no difference in age, gender, height of fever, asthma diagnosis, presence of rhinorrhea, congestion, grunting, or shortness of breath in children diagnosed with ACS compared to those with a negative CXR. Clinical characteristics involving patients discharged from the ED and <72-h ED returns are summarized in Supporting Information Methods/Results. One patient who was discharged without a CXR returned to the ED within 24 h with ACS. Figure 1A summarizes multivariate logistic regression analysis, identifying history of ACS, cough, chest pain, and abnormal PE findings as independent risk factors associated with ACS, with an area under a receiver operating characteristics curve (AUC) of 0.7254, 95% CI [0.6612–0.7893]. Multivariate logistic regression was also applied to the 2014 NIH guidelines (Fig. 1A). There is a statistically significant difference in AUC between Emory/CHOA and the NIH risk stratification model, with NIH guidelines giving a lower AUC = 0.6623, 95% CI [0.5946–0.7299] (P = 0.04; Fig. 1B). Of the 66 CXRs found to be positive, no cases of ACS occurred in children who lacked all Emory/CHOA risk factors. Six out of 66 (9.1%) children with a positive CXR had a history of ACS alone and no other risk factors. Percentage of patients with ACS presenting with 0–4 risk factors is summarized in Supporting Information Table III. Of the 230 children with SCD and fever who were not evaluated by CXR, 112 (48.7%) had at least one of our identified independent risk factors for ACS. (A) Multivariate logistic regression analysis with adjusted point estimates and 95% confidence intervals (CI) for odds ratios and (B) area under the receiver operating characteristic curve (AUC) comparison of the diagnostic performance of National Institutes of Health (NIH) fever guidelines versus Emory/Children's Healthcare of Atlanta (CHOA) proposed acute chest syndrome (ACS) risk factor model. Multivariate regression analysis included a history of ACS, chest pain, cough or abnormal physical exam findings as risk factors in our model to identify ACS, with an AUC = 0.7254. No children lacking all of these risk factors were found to have ACS. This prediction model performed better than the NIH guidelines for ACS risk (AUC = 0.6623, P = 0.04). However, both models demonstrated poor accuracy to identify ACS. Supporting Information Table IV summarizes data from 29 children who had multiple ED encounters with at least 1 positive and 1 negative CXR and were subsampled to perform a matched comparison. Our goal was to identify risk factors for ACS that might be helpful for the evaluating ED physician in order to identify children requiring a CXR as part of their fever work-up. These data reveal both high and low ACS-risk models that include some unique features not previously described. The AUC for NIH-criteria risk factors when applied to our data set to identify ACS was only 0.6623. The Emory/CHOA model of a history of ACS, chest pain, cough, or abnormal physical exam findings appear to be better predictors of ACS with an AUC of 0.7254. There was a statistically significant difference between the 2 AUC, with our risk factors being more sensitive and specific than the current NIH guidelines to identify cases of ACS in this large cohort of children with SCD and fever. It may be useful to add “history of ACS” as one of the recommended criteria to obtain a CXR during a fever work-up. However, an AUC < 0.80 is still insufficient to accurately identify all ACS and is not an ideal prediction model. Although there are clinical signs and symptoms that identify ACS risk 1, 2, 5, the majority of patients will present early on with a normal lung exam, requiring a high level of suspicion given the morbidity and mortality associated with ACS. The results of this study mirror the findings of previous studies performed nearly 20 years ago that address this issue 1, 2. Despite recent advances in medicine, many children with SCD and fever still have few to no additional symptoms to suggest ACS when it is present. Clinical assessment alone is not reliable in identifying ACS, since 62% of children with ACS in our large cohort had a normal lung exam. Although the desire to decrease diagnostic radiation is an admirable goal 6, the SCD population may not be the ideal group in whom radiographs should be limited due to the high morbidity and mortality of ACS. Obtaining a CXR should be strongly considered in any child with SCD/fever and a history of ACS, chest pain or any respiratory symptoms, as missing ACS can lead to significant morbidity and mortality. Sarah G. Lazarus,1,2 Michael Kelleman,3 Olufolake Adisa,2,3,4 April R. Zmitrovich,2 Robert Hagbom,5 Stephanie Cohen,2,3,6 Courtney E. McCracken,3 and Claudia R. Morris 2,3,6,7* 1 Pediatric Emergency Medicine Associates (PEMA-LLC), Atlanta, Georgia; 2 Children's Healthcare of Atlanta, Atlanta, Georgia; 3 Department of Pediatrics, Emory University School of Medicine, Atlanta, Georgia; 4 Children's Healthcare of Atlanta, AFLAC Cancer and Blood Disorders Center, Atlanta, Georgia; 5 Department of Pediatrics, University of Hawaii, Honolulu, Hawaii; 6 Division of Pediatric Emergency Medicine, Emory University School of Medicine, Atlanta, Georgia; 7 Emory-Children's Center for Cystic Fibrosis and Airways Disease Research, Emory University School of Medicine, Atlanta, Georgia Additional Supporting Information may be found in the online version of this article. Supporting Information Supporting Information Supporting Information Supporting Information Supporting Information Supporting Information Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.