Trends in Use of Advanced Imaging in Pediatric Emergency Departments, 2009-2018

Key Points

Question  How has the use of advanced imaging changed during the past 10 years across pediatric emergency departments in the US?

Findings  This cross-sectional study of 26 million pediatric emergency department visits revealed that, overall, the rate of advanced imaging increased from 6.4% of encounters in 2009 to 8.7% in 2018; specifically, the rate of computed tomography decreased from 3.9% to 2.9%, the rate of ultrasonography increased from 2.5% to 5.8%, and the rate of magnetic resonance imaging increased from 0.3% to 0.6%.

Meaning  Overall use of advanced imaging has increased during the past decade, driven by the use of nonradiating modalities replacing the use of computed tomography.

Importance  There is increased awareness of radiation risks from computed tomography (CT) in pediatric patients. In emergency departments (EDs), evidence-based guidelines, improvements in imaging technology, and availability of nonradiating modalities have potentially reduced CT use.

Objective  To evaluate changes over time and hospital variation in advanced imaging use.

Design, Setting, and Participants  This cross-sectional study assessed 26 082 062 ED visits by children younger than 18 years from the Pediatric Health Information System administrative database from January 1, 2009, through December 31, 2018.

Exposures  Imaging.

Main Outcomes and Measures  The primary outcome was the change in CT, ultrasonography, and magnetic resonance imaging (MRI) rates from January 1, 2009, to December 31, 2018. Imaging for specific diagnoses was examined using all patient-refined diagnosis related groups. Secondary outcomes were hospital admission and 3-day ED revisit rates and ED length of stay.

Results  There were a total of 26 082 062 visits by 9 868 406 children (mean [SD] age, 5.59 [5.15] years; 13 842 567 [53.1%] male; 9 273 181 [35.6%] non-Hispanic white) to 32 US pediatric EDs during the 10-year study period, with 1 or more advanced imaging studies used in 1 919 283 encounters (7.4%). The proportion of ED encounters with any advanced imaging increased from 6.4% (95% CI, 6.2%-6.2%) in 2009 to 8.7% (95% CI, 8.7%-8.8%) in 2018. The proportion of ED encounters with CT decreased from 3.9% (95% CI, 3.9%-3.9%) to 2.9% (95% CI, 2.9%-3.0%) (P < .001 for trend), with ultrasonography increased from 2.5% (95% CI, 2.5%-2.6%) to 5.8% (95% CI, 5.8%-5.9%) (P < .001 for trend), and with MRI increased from 0.3% (95% CI, 0.3%-0.4%) to 0.6% (95% CI, 0.6%-0.6%) (P < .001 for trend). The largest decreases in CT rates were for concussion (−23.0%), appendectomy (−14.9%), ventricular shunt procedures (−13.3%), and headaches (−12.4%). Factors associated with increased use of nonradiating imaging modalities included ultrasonography for abdominal pain (20.3%) and appendectomy (42.5%) and MRI for ventricular shunt procedures (17.9%) (P < .001 for trend). Across the study period, EDs varied widely in the use of ultrasonography for appendectomy (median, 57.5%; interquartile range [IQR], 40.4%-69.8%) and MRI (median, 15.8%; IQR, 8.3%-35.1%) and CT (median, 69.5%; IQR, 54.5%-76.4%) for ventricular shunt procedures. Overall, ED length of stay did not change, and hospitalization and 3-day ED revisit rates decreased during the study period.

Conclusions and Relevance  This study found that use of advanced imaging increased from 2009 to 2018. Although CT use decreased, this decrease was accompanied by a greater increase in the use of ultrasonography and MRI. There appears to be substantial variation in practice and a need to standardize imaging practices.

Computed tomography (CT), ultrasonography, and magnetic resonance imaging (MRI) are important diagnostic tools in the evaluation of children seeking emergency department (ED) care. Historically, CT has been the most commonly used advanced imaging modality in the ED setting¹ because it is fast, accurate, and accessible. However, scientific consensus is that radiation from CT is associated with increased risk of cancer, particularly with exposure in childhood.² Advocacy campaigns, such as Image Gently² and Choosing Wisely,³ and media reports,^4-7 have raised awareness in the medical community and the public about these risks. Consequently, health care professionals, patients, and families may seek alternative, nonradiating modalities, such as ultrasonography and MRI, when they are considered effective and it is feasible. The benefits of ultrasonography and MRI are balanced by the often limited availability, operator dependence, cost, and, in some cases for MRI, the need for sedation.

Prior studies^8-13 evaluating ED imaging use have focused on specific diagnoses or local patterns of use or have only examined use of CT without considering alternative imaging options. Ultrasonography has increasingly been used as a replacement for the evaluation of several conditions that traditionally have relied on CT.^14-16 Advances in MRI technology, including the development of so-called fast or rapid protocols,^17-19 and increased availability have led to greater feasibility of MRI for children in the ED. In addition, clinical decision rules^20,21 and objective scoring tools^22-24 have provided opportunities to safely minimize unnecessary imaging altogether. It is for these reasons that emergency imaging patterns have likely continued to change. A comprehensive, up-to-date assessment of pediatric ED-specific imaging practices and patterns can provide benchmarks for practitioners, hospitals, and insurers. Thus, the primary objective of this study was to evaluate the changes in the use of advanced imaging, including diagnosis-specific patterns, in US-based pediatric EDs between 2009 and 2018. We hypothesized that CT use decreased with a concomitant increase in ultrasonography and MRI. Our secondary objectives were to assess variation in these imaging patterns across pediatric EDs and to examine changes in hospitalization and 3-day revisit rates and ED length of stay (LOS).

We conducted a multicenter cross-sectional study using the Pediatric Health Information System (PHIS), an administrative database that includes ED and inpatient data from 52 tertiary care children’s hospitals in the US and is housed by the Children’s Hospital Association. Data quality is assured through a joint effort between the Children’s Hospital Association and participating hospitals. Data are deidentified and subject to multiple validity and reliability checks before inclusion in the database. For this study, we included 32 EDs in the final cohort after excluding 20 hospitals that did not contribute comprehensive ED-specific data during the study period. We included visits by patients younger than 18 years with an ED encounter from January 1, 2009, through December 31, 2018. The University of Pittsburgh Human Research Protection Office determined the study was not human subjects research and was therefore exempt from review and informed consent.

The primary outcome was the proportion of ED encounters with at least 1 advanced imaging study, identified through billing codes, and defined as CT, ultrasonography, MRI, or a combination of these. The database permits identification of the date the imaging was performed; however, it does not distinguish between ED and inpatient testing for patients admitted from the ED. Therefore, for patients admitted to the hospital from the ED, we defined ED imaging based on time of arrival. Specifically, we attributed imaging to the ED visit when the ED arrival time was before 6 pm and on the initial or second hospital day for an admitted patient with an ED arrival after 6 pm.²⁵ For children who were discharged from the ED, we examined their entire ED encounter.

In addition to evaluating overall use of advanced imaging, as a subanalysis, we focused on conditions that were associated with the overall decreases in CT; specifically, we evaluated conditions associated with CT changes above the mean decrease, comparing 2018 with 2009. Encounter diagnoses were categorized using the all patient-refined diagnosis related groups (APR-DRGs), a standardized method of classifying patients into 1 of 311 clinically relevant groups, according to International Classification of Diseases (ICD) diagnosis codes, severity of illness, and mortality risk.²⁶ To avoid including APR-DRGs with low absolute frequencies of imaging, for this analysis, we excluded diagnoses with a mean of fewer than 30 CT scans performed per hospital annually.²⁷

To examine potential unintended consequences associated with changes in advanced imaging, for these conditions, we assessed 3 secondary outcome measures: (1) ED LOS for discharged patients, (2) hospitalization rates, and (3) 3-day ED revisit rates. For measuring LOS, we examined only patients discharged from the ED, excluding hospitalized patients because it is not possible in the database to isolate ED-specific LOS for admitted patients. We also assessed patient characteristics across the study period, including age, sex, race/ethnicity, presence of a complex chronic condition²⁸ during the encounter, primary insurance, ED disposition, and geographic region.

We evaluated overall and APR-DRG–specific annual trends of each advanced imaging modality as a proportion of ED visits. We summarized categorical variables using frequencies and percentages and continuous variables with medians and interquartile ranges (IQRs). We used χ² tests to compare categorical data. For the analysis of CT trends, we used nonlinear modeling optimization to identify the year in which there was a change in slope. We used generalized estimating equations, which accounted for hospital clustering, to analyze trends over time. We constructed groups of generalized linear models with a random effect for hospital, evaluating the independent association of relevant covariates with our primary outcomes of modes of advanced imaging (first group) and with secondary outcome measures of ED LOS, hospitalization, and revisit rates (second group). Significance was set at 2-sided P < .05. All analyses were performed using SAS software, version 9.4 (SAS Institute Inc).

There were a total of 26 082 062 visits to 32 US pediatric EDs during the 10-year study period, with 1 or more advanced imaging studies used in 1 919 283 encounters (7.4%). A total of 9 273 181 patients (35.6%; mean [SD] age, 5.59 [5.15] years; 13 842 567 [53.1%] male) were non-Hispanic white, 7 136 412 (27.4%) were non-Hispanic black, and 7 543 194 (28.9%) were Hispanic (Table). A total of 16 511 436 (64.4%) had public insurance, and most 22 455 390 (86.1%) were discharged from the ED. No substantive changes in patient demographics were found across the study period.

The percentage of ED encounters associated with any advanced imaging increased from 6.4% (95% CI, 6.4%-6.4%) in 2009 to 8.7% (95% CI, 8.7%-8.8%) in 2018 (P < .001 for trend) (Figure 1), resulting in 65 613 more encounters with advanced imaging in 2018. Computed tomography use decreased from 3.9% (95% CI, 3.9%-3.9%) in 2009 to 2.9% (95% CI, 2.9%-3.0%) in 2018 (P = .001 for trend), resulting in 27 487 fewer encounters with CT in 2018. Despite this decrease during the 10-year period, there was a plateau from 2014 to 2018. Specifically, CT use decreased from 3.9% (95% CI, 3.9%-3.9%) to 2.9% (95% CI, 2.8%-2.9%) from 2009 to 2014 (P < .001) but did not change from 2014 (2.9%; 95% CI, 2.8%-2.9%) to 2018 (2.9%; 95% CI, 2.7%-2.9%) (P = .32). Ultrasonography use increased from 2.5% (95% CI, 2.5%-2.6%) in 2009 to 5.8% (95% CI, 5.8%-5.9%) in 2018 (P < .001 for trend), resulting in 92 852 more encounters with ultrasonography in 2018, and MRI use increased from 0.3% (95% CI, 0.3%-0.4%) in 2009 to 0.6% (95% CI, 0.6%-0.6%) in 2018 (P < .001 for trend), resulting in 8116 more encounters with MRI in 2018. There were similar patterns of increases in use of MRI and ultrasonography and decreases in use of CT when stratifying encounters by ED disposition (Figure 1 and eTable 1 in the Supplement). Of the 1 919 283 ED encounters with advanced imaging, 133 862 (7.0%) included more than 1 advanced imaging modality, and 92 232 (68.9%) of these encounters included both ultrasonography and CT. Children 5 years and older, white patients, those with complex chronic conditions, those with private insurance, and those who were admitted had higher odds of undergoing CT, MRI, and ultrasonography (eTable 2 in the Supplement).

There were 19 APR-DRGs associated with at least 30 CT encounters per hospital per year. The percentage change (median, −2.1%) in CT across the study period ranged from −23.0% to 0.9% across these 19 APR-DRGs. Among these conditions, there were 8 APR-DRGs with decreases in use of CT that were greater than the mean (−5.2%; 95% CI, −8.1% to −2.3%) decrease, which were the focus of this subanalysis: concussion, closed skull fracture, uncomplicated intracranial, coma of less than 1 hour, or no coma (−23.0% change; 95% CI, −24.1% to −21.9%); appendectomy (−14.9% change; 95% CI, −16.0% to −13.8%); ventricular shunt procedures (−13.3% change; 95% CI, −16.7% to −10.0%); migraine and other headaches (−12.4% change; 95% CI, −13.0% to −11.7%); other disorders of the nervous system (−10.1% change; 95% CI, −11.3% to −8.9%); abdominal pain (−6.1% change; 95% CI, −6.4% to −5.8%); other ear, nose, mouth, throat, and cranial or facial diagnoses (−5.9% change; 95% CI, −6.2% to −5.7%); and seizure (−5.3% change; 95% CI, −5.8% to −4.8%) (P < .001 for trend). These APR-DRGs were associated with 51.1% of CT imaging and 31.5% of advanced imaging performed during the study period. APR-DRG–specific patterns of CT, ultrasonography, and MRI use over time are shown in Figure 2, Figure 3, and eTable 3 in the Supplement. These APR-DRGs were associated with more than 1 advanced imaging modality in 14.7% (ventricular shunt procedures), 8.6% (appendectomy), 3.4% (abdominal pain), 3.1% (other disorders of the nervous system), 1.2% (seizure), 0.9% (migraine), 0.8% (concussion), and 0.1% (other ear, nose, mouth, throat, and cranial or facial diagnoses) of encounters. The most common principal ICD-9 or ICD-10 diagnosis codes for each of these 8 APR-DRGs are listed in eTable 4 in the Supplement.

Across the 8 APR-DRGs, 4 conditions were associated with an overall increase in advanced imaging use despite the decrease in CT use. Specifically, this increase from 2009 to 2018 was greatest in ultrasonography for appendectomy (42.5%; 95% CI, 41.3%-43.8%) and abdominal pain (20.3%; 95% CI, 19.8%-20.8%) and MRI for ventricular shunt procedures (17.9%; 95% CI, 15.0%-20.8%) (P < .001 for trend). For other conditions, most notably concussion, the decrease in CT use was not associated with a compensatory increase in other modalities.

Variability of advanced imaging across EDs for the 8 APR-DRGs is shown in Figure 4. We observed the widest variation across the 32 EDs in the use of ultrasonography for appendectomy (median, 57.5%; IQR, 40.4%-69.8%) and MRI (median, 15.8%; IQR, 8.3%-35.1%) and CT (median, 69.5%; IQR, 54.5%-76.4%) for ventricular shunt procedures. Emergency department–specific changes in CT use, comparing the years before and after the plateau in CT in 2014, are shown in eFigure 1 in the Supplement. For encounters among the 8 APR-DRGs, the median baseline CT use across EDs was 15.0% (range, 8.3%-27.6%), with rates changing across the 2 periods by −10.7% before 2014 to 1.4% after 2014. The EDs with baseline CT rates in the lowest quartile did not have large changes in their overall rates of CT. Four hospitals consistently had reduced CT use greater than the median decrease for all conditions, whereas others did so only for certain conditions.

Across the 8 APR-DRGs, advanced imaging was independently associated with an increase in ED LOS and odds of admission and 3-day revisits (eTable 5 in the Supplement). However, despite these encounter-level increases and an increase in advanced imaging across the study period, overall ED LOS remained stable across the study period, with decreasing rates of admission and 3-day revisits (eFigure 2 in the Supplement).

We observed an overall increase in the use of advanced imaging across US pediatric EDs during the past decade. The use of CT decreased, with most of this decrease associated with 8 APR-DRGs. For certain APR-DRGs, this decrease was associated with increases in other radiation-sparing advanced imaging modalities, such as ultrasonography for abdominal conditions and MRI for patients with ventricular shunts. The magnitude of the decrease in CT use for these conditions varied by hospital but was not associated with increases in hospitalizations, ED revisits, or ED LOS.

Reasons for these changing imaging patterns are likely multifactorial. Epidemiologic studies^29-31 have provided more insight for the scientific community into the association between radiation exposure from CT and cancer risk. There have been several high-profile media publications^7,32 that have likely provoked concerns regarding CT from patients and families.³³ Campaigns that target health care professionals and the public, such as Image Gently in 2007² and Choosing Wisely in 2012,³⁴ have raised awareness regarding optimizing pediatric imaging practices. Efforts such as these may, in part, explain the trends toward nonradiating imaging modalities.

For example, the American Academy of Pediatrics³⁵ and the American College of Radiology³⁶ Choosing Wisely lists of “ten things physicians and patients should question” recommend that practitioners consider performing ultrasonography first before or in lieu of CT for children with abdominal pain. Although the diagnostic accuracy of CT is superior to ultrasonography for appendicitis,³⁷ EDs have used various strategies to reduce CT use and thus radiation exposure for the evaluation of abdominal pain. Several studies^38-41 highlight the association of guideline implementation and integration of clinical decision support into electronic health systems with reducing abdominal CT use without increasing adverse outcomes. Such strategies may be associated with some of the increase in ultrasonography use we observed in patients with diagnoses related to abdominal issues (ie, abdominal pain and appendectomy). Interestingly, the increase in the use of ultrasonography for these conditions was greater than the decrease in the use of CT. Therefore, a proportion of these ultrasonographic examinations likely represent overuse of a typically widely available and nonradiating imaging tool. With respect to concussion encounters, the reduction in CT use was not supplanted by increases in alternative imaging modalities. The decreases in CT use in both instances may reflect the knowledge translation of clinical decision rules,^{20,22-24,42,43} whereas the divergence in replacement may reflect alternative options: ultrasonography for appendicitis is relatively inexpensive, readily available, and safe, whereas MRI for intracranial hemorrhage is expensive and less available and, although safe, often requires sedation.

Another area of imaging growth we observed in our study was the use of MRI in patients with ventricular shunt procedures. In the past decade, studies^17,44-47 have highlighted the accuracy and feasibility of so-called rapid or fast MRI in lieu of CT for the evaluation of ventricular shunt malfunction. With rapid MRI technology, there is often no requirement for sedation, even for infants and young children. As more indications for these rapid MRI studies emerge and MRI availability increases,^18,19,48-50 MRI adoption may continue to increase and may be associated with further decreases in CT imaging for many diagnoses.⁵¹

For certain APR-DRGs, specifically concussion, the decrease in CT use was not associated with an increase in other advanced imaging modalities. Clinical decision rules for head trauma^20,42,43 allow for the safe selection of low-risk patients who can be observed without neuroimaging. These decreases in CT may be associated with dissemination and implementation⁵² of these clinical decision rules. It is also plausible that given the increases in awareness and attention surrounding concussion in the past decade,⁵³ the volume and recognition of patients with concussion have increased, resulting in a lower overall proportion of CT in this population.

Our findings build on other studies^9-11,54,55 of imaging in children. Specifically, earlier work^9-11,54 demonstrated decreases in CT use among pediatric ED patients and increases in ultrasonography and MRI use for certain conditions beginning in the mid-2000s. Our work provides additional insight into more recent trends, revealing an unexpected plateau in CT use since 2014. This finding may indicate that, at least with current efforts, a lower limit to CT use has been reached. A recent study by Smith-Bindman et al⁵⁵ examined trends in advanced imaging and found increases in ultrasonography beyond the decreases in CT use, as well as a similar plateau after the more rapid decreases in CT use. That study, however, examined encounters by privately insured patients across all health care settings. Our study examined ED-specific practices and allowed for more detailed investigation into specific diagnoses associated with imaging trends over time.

We observed significant between-site variation, which may be secondary to local guidelines and protocols, equipment availability, and practice preferences among subspecialists.⁵⁶ Our findings provide benchmarks for EDs that seek to understand their imaging practices in the context of peer hospitals. Further reductions in CT use at some sites remain possible but likely require a greater emphasis on knowledge translation and quality improvement initiatives.^38,52,57

Our study has several limitations. We were unable to assess the indications for imaging studies, and unmeasured confounders related to the performance of advanced imaging may exist that could have affected our findings. Similarly, we do not know if there were harms, such as incidental findings that required additional visits, more downstream testing, and direct costs to patients and the health care system. Some revisits were not captured in our database because, although the PHIS database allows for tracking individual patients longitudinally, only revisits to the same hospital are captured. If patients returned to other EDs for additional care, we would have underestimated the frequency of 3-day ED revisits. The effect of this limitation, however, is likely to be small because patients are more likely to return to the index ED.⁵⁸ APR-DRGs are assigned using diagnostic algorithms that are based on discharge codes associated with individual encounters.²⁶ In addition, severity of illness is ascribed to the encounter based on multiple factors, including the use of diagnostic imaging. Therefore, some patients may be misclassified when assigned to certain APR-DRGs (eg, a patient with a clinically insignificant intracranial hemorrhage seen on CT may be assigned a different APR-DRG than if a CT had not been performed and the patient was diagnosed with concussion). This study was conducted across children’s hospitals in the US and may not be generalizable to other hospitals or institutions in other countries. However, the trends we observed in rates of CT, ultrasonography, and MRI in children’s hospital EDs mirror those found in the more comprehensive analysis by Smith-Bindman et al.⁵⁵ We may have misclassified some imaging studies as having occurred within the ED for hospitalized patients who had imaging after admission to the hospital. In addition, point-of-care ultrasonographic examinations by emergency physicians may not have been captured because PHIS data are based on procedures billed and site variation in billing exists for these examinations.⁵⁹ This occurrence would have resulted in an underestimation of true ultrasonography use rates.

The study found that overall use of advanced imaging in pediatric EDs in the US increased between 2009 and 2018. A decrease in CT use occurred during the 10-year period, with a plateau after 2014. The magnitude of the decrease in CT use varied by site and was associated, in part, with the use of alternative advanced imaging modalities for certain diagnoses, most notably ultrasonography for abdominal conditions. Future efforts appear to be needed to standardize imaging approaches and evaluate the effect of the changing landscape of advanced imaging on patient-level outcomes.

Accepted for Publication: March 29, 2020.

Corresponding Author: Jennifer R. Marin, MD, MSc, Division of Pediatric Emergency Medicine, UPMC Children's Hospital of Pittsburgh, 4401 Penn Ave, Administration Office Building, Ste 2400, Pittsburgh, PA 15224 (jennifer.marin@chp.edu).

Published Online: August 3, 2020. doi:10.1001/jamapediatrics.2020.2209

Author Contributions: Dr Marin and Mr Rodean had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Marin, Rodean, Hall, Chaudhari, Cohen, Freedman, Peltz, Shah, Simon, Neuman.

Acquisition, analysis, or interpretation of data: Marin, Rodean, Hall, Alpern, Aronson, Chaudhari, Cohen, Freedman, Morse, Samuels-Kalow, Shah, Simon, Neuman.

Drafting of the manuscript: Marin, Hall, Morse, Simon, Neuman.

Critical revision of the manuscript for important intellectual content: Marin, Rodean, Alpern, Aronson, Chaudhari, Cohen, Freedman, Morse, Peltz, Samuels-Kalow, Shah, Simon, Neuman.

Statistical analysis: Rodean, Hall.

Administrative, technical, or material support: Freedman, Peltz, Neuman.

Supervision: Marin, Chaudhari, Freedman, Simon, Neuman.

Conflict of Interest Disclosures: Dr Aronson reported receiving grant support from the Agency for Healthcare Research and Quality during the conduct of the study. Dr Freedman reported receiving grant support from the Alberta Children’s Hospital Foundation Professorship in Child Health and Wellness. Dr Samuels-Kalow reported receiving a career development grant from the Harvard Catalyst–Harvard Clinical and Translational Science Center (National Center for Advancing Translational Sciences, National Institutes of Health) and financial contributions from Harvard University and its affiliated academic health care centers. No other disclosures were reported.

Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Healthcare Research and Quality, the Alberta Children's Hospital Foundation, Harvard Catalyst, Harvard University and its affiliated academic health care centers, or the National Institutes of Health.