Investigation of Chest X-Ray Use in the Emergency Department in Pediatric Patients with Sickle Cell Disease Presenting with Fever Compared to Age Matched Controls

Background

In sickle cell disease (SCD), acute chest syndrome (ACS) is associated with prolonged hospitalization, increased risk of respiratory failure, future lung disease and 25% mortality in hospitalized patients (Bakshi, & Krishnamurti, 2017; Vinchinsky et al. 1997). Pediatric patients with SCD frequently present to the Pediatric Emergency Department (PED) with complaints of fever, chest pain, and cough, all of which may or may not be related to ACS. It is challenging for PED providers to determine which patients are at highest risk of ACS, so chest X-Rays (CXR) are frequently ordered which increases radiation exposure and healthcare costs. The objective of this study was to identify incidence of CXR performance, as well as ACS diagnosis, in SCD patients presenting to our PED with or without fever. Our goal was to identify significant clinical predictors of ACS in this population in order to implement a diagnostic algorithm for PED providers.

Methods

This was an IRB-approved retrospective medical record review of subjects diagnosed with SCD with inclusion criteria: ages 2-12 years, who presented to the University of Maryland PED between 2016-2018. We performed bivariate analyses comparing these variables between subjects who were febrile vs. afebrile on presentation to the PED, as well as those who were ultimately diagnosed with ACS compared to those who were not. Analysis of categorical variables was performed using Chi-square or Fischer exact test as appropriate. We performed a multivariable logistic regression model to identify significant predictors of ACS diagnosis. Analyses performed using SAS 9.4.

Results

We identified 424 SCD subjects who presented to our PED meeting inclusion criteria, with 25% (n=108) presenting with fever. Of these, 69% received a CXR on presentation vs. 42% of afebrile subjects (p=<0.0001). In our febrile group 21% (n=23) patients had more than 2 febrile episodes and 100% received CXRs. There were no significant differences between the febrile and afebrile subjects when it came to sex, asthma diagnosis/comorbidity, hydroxyurea use, folic acid supplementation, or pneumococcal prophylaxis. Overall, 10% of patients presenting to the PED were diagnosed with ACS (n=42), made up of 13% of those presenting with fever vs. 9% of those presenting without fever. Those subjects ultimately diagnosed with ACS were significantly more likely to present with chest pain (p=0.003), tachypnea (p=0.001), and hypoxia (p<0.0001), and were more likely to have a past history of asthma (p=0.0085). Sickle cell variant, home medications, and history of splenectomy were not significantly associated with ACS diagnosis. Upon multivariable modeling, when adjusting for fever and pre-existing asthma diagnosis, the only significant predictors of ACS diagnosis were chest pain and hypoxia. Patients without chest pain had an odds ratio (OR) =0.3 of ACS diagnosis [95% Confidence Interval, CI 0.14-0.67], indicating they had 70% lower odds of ACS compared to patients with chest paint. Patients without hypoxia had OR=0.12 of ACS compared to those with hypoxia [CI 0.06-0.25], indicating an 88% reduced odds of ACS diagnosis. Conversely, those with chest pain had 3.3x the odds of ACS diagnosis [CI 1.5-7.4] and those with hypoxia had 8.4x the odds of ACS diagnosis [CI 4-17.9] compared to those without these symptoms.

Conclusion

In ACS, current guidelines recommend that patients presenting with fever, hypoxia, tachypnea, tachycardia and abnormal respiratory exam findings should be treated empirically as well as receive a CXR. However radiological signs can be delayed compared to physical signs so a normal CXR does not preclude the diagnosis of ACS if there is clinical suspicion (Howard et al. 2015). Our data demonstrate that clinical findings such as chest pain, tachypnea and hypoxia were most likely to correlate to a diagnosis of ACS. While 69% of our febrile patients received a CXR in the PED, only 13% were ultimately diagnosed with ACS, indicating that more CXRs and radiation exposure occurred in the febrile population than may have been necessary. When adjusting for fever and asthma, the most notable predictors of ACS were hypoxia and chest pain. When present, these findings are significant predictors of ACS; when absent, subjects had significantly decreased odds of ACS. Incorporating the presence or absence of chest pain and hypoxia may help focus the use of CXR on the appropriate patient population.