U.S. Diagnostic Reference Levels and Achievable Doses for 10 Pediatric CT Examinations

Introduction

Diagnostic reference level (DRL) benchmarks for radiation protection and optimization of patient imaging were first mentioned by the International Commission on Radiological Protection (ICRP) in 1990 (1) and clarified further in 1996 (2). The DRL is used for identifying situations where the levels of patient dose are unusually high (2,3). The use of DRLs is endorsed by professional, advisory, and regulatory organizations. The ICRP emphasizes that DRL values “are not for regulatory or commercial purposes, not a dose restraint and not linked to limits or constraints” (4). DRLs are typically set at the 75th percentile of the dose distribution from a survey conducted across a broad user base using a specified dose measurement protocol. They are established both regionally and nationally, and considerable variations have been seen across regions and countries (4–6)

The concept of achievable dose (AD) was introduced in 1999 in a United Kingdom National Radiation Protection Board advisory group to recognize more typical values within a practice (7). In 2012, the National Council on Radiation Protection and Measurements proposed that ADs be set at the median, or 50th percentile, of a dose survey, on the basis that 50% of the facilities have already achieved doses at or below this value (8).

In 2017, DRLs were published for adult CT examinations in the United States using the American College of Radiology CT Dose Index Registry (DIR) (9–11). Few current U.S. recommendations exist for pediatric DRLs and ADs in CT. Some are based on phantom surveys, not clinical examinations (8,12). Other work in the United States has used clinical examinations from DIR data to provide age- or size-based dose index guidance for pediatric CT (13–15). In 2018, the European Commission summarized existing national DRLs from 17 countries in its Guidelines for Diagnostic Reference Levels for Pediatric Imaging based mainly on the anatomic region imaged; these are primarily presented in age groups and, in some cases, weight groupings (16). At this time, there are no national DRLs based on patient dimensions or size. The purpose of this study is to use the CT DIR to develop robust, current, national DRLs and ADs for the 10 most commonly performed pediatric CT examinations as a function of patient age and size. This is the first time these benchmarks have been developed in the United States.

Materials and Methods

The 10 most commonly performed CT examinations in the United States performed between January 2016 and December 2020 in patients aged 18 years and younger were included in the study as follows: (a) head without contrast material, (b) sinuses without contrast material, (c) maxillofacial without contrast material, (d) neck with contrast material, (e) cervical spine without contrast material, (f) chest without contrast material, (g) chest with contrast material, (h) abdomen and pelvis without contrast material, (i) abdomen and pelvis with contrast material, and (j) chest, abdomen, and pelvis with contrast material.

The DIR uses the RSNA RadLex^® Playbook identifier (17) for mapping the CT examinations (Table 1). Only examinations from facilities inside the United States with complete patient age and size information, dose indexes, and study descriptions were included. No patient-identifiable information was sent to the American College of Radiology. Research conducted on data from any of the American College of Radiology registries is exempt from institutional review board approval. Multiphase examinations (ie, both non-contrast and contrast-enhanced examinations), body examinations with missing effective diameter, and head examinations with missing age were excluded to prevent misrepresentation of the radiation dose. DIR also has a process for identifying multiregion CT examinations, thus avoiding these examinations from being included incorrectly in the analysis. If an examination is conducted using one irradiation event to satisfy multiple orderables, the DIR identifies the split examination by comparing the examination metadata, such as study time, patient date of birth, and other parameters, to determine exact matches. These matched examination records are combined into combination examinations in the DIR for the participants to map accordingly. This process was validated and found to be accurate (18). Examinations from age and size bins with fewer than five facilities and facilities reporting fewer than 20 examinations of a particular examination type during the time period of the study were also excluded.

Body Examinations

Body examinations were characterized in terms of effective diameter (19). Although effective diameter is not displayed on the CT console, it can be determined from patient dimensions or from the dose monitoring software program that may display these dimensions. In this study, the effective diameter was automatically determined from the anteroposterior and lateral dimensions measured from localizer images using a size estimation algorithm (20) following American Association of Physicists in Medicine methods (21). Since the DIR does not collect axial slice data, it relies on patient size estimation from the localizer. We used the algorithm from Duke University for this purpose (20). The process is robust across most body parts with unique challenges in the case of the head and neck because of the inconsistent inclusion of head and shoulders across patients, particularly pediatric and bariatric patients, where the neck presents limited axial extent. As patient size is taken as the median patient thickness from each row of the localizer, the neck measurement becomes highly dependent on the amount of shoulder present in the localizer and thus unreliable from either anteroposterior or lateral localizers. The head measurements likewise present a different challenge as the nose and soft tissue of the face inconsistently contribute to the estimate in the lateral views. This is due to the opacity of these tissues relative to the rest of the head, which sometimes places the intensity of the nose and facial soft tissue in lateral scout images below intensity thresholds used for image segmentation. Thus, for the head and neck examinations, only age was used to stratify the data. Although size-specific dose estimate (SSDE) conversion factors for head examinations were published by the American Association of Physicists in Medicine in July 2019 (22), most of the data in this analysis were acquired before July 2019, and as such, head SSDE was not calculated for this study and was deferred to future integration into the DIR. For body examinations, the effective diameter was used to determine the appropriate conversion factor to estimate SSDE from volume CT dose index (CTDI_vol) normalized to a 32-cm phantom.

International guidance was used when setting groupings for pediatric age- and size-based analyses when possible. Head examination age groups were partially based on European guidelines (16). The European age groups of 0 to less than 3 months and 3 months to less than 1 year were combined because the DIR does not collect age information in months. Also, the age group of 1 to less than 6 years was split into two groups because of the substantial changes in head size during that age range. Head, sinus, and maxillofacial examinations for this study were grouped into the following age bins: (a) 0 to less than 1 year, (b) 1 to less than 2 years, (c) 2 to less than 6 years, and (d) 6–18 years. The age-based bins for body examinations were based on the ICRP 135 bands of 0, 1, 5, 10, and 15 years (6). Body examinations were therefore grouped into the following age bins: (a) 0 to less than 1 year, (b) 1 to less than 5 years, (c) 5 to less than 10 years, (d) 10 to less than 15 years, and (e) 15–18 years. There are no international recommendations on grouping pediatric neck studies, so the body age groupings were used.

International size-based grouping recommendations are based on weight. The DIR does not collect weight information but provides effective diameter, which is more directly applicable to the determination of SSDE. Although there are pediatric bin-size precedents for dose versus effective diameter (14,23), the larger patient population in this study allowed us to extend the upper range beyond those previously published to include effective diameters for larger, teenage pediatric patients and those that may be considered obese. To allow for comparison of the new pediatric DRLs with the 2017 adult DRLs (9), pediatric bins were set analogous to those generally used in the adult publication as follows: (a) 0 to less than 12 cm, (b) 12 to less than 16 cm, (c) 16 to less than 20 cm, (d) 20 to less than 24 cm, (e) 24 to less than 28 cm, (f) 28 to less than 32 cm, (g) 32 to less than 36 cm, (h) 36 to less than 40 cm, and (i) more than 40 cm.

Statistical Analysis

Descriptive statistics were calculated for facility category, location, census region, average volume of examinations per month, patient age, and sex. One-way frequency tables were generated for demographic distributions of the study population, which included frequency, number of facilities, and number of examinations. To allow for some comparison with DRLs published in other countries, international guidance was followed in analyzing data where possible. ICRP 135 (6) specifically defines national DRL values as “the third quartile (75th percentile) of the distribution of the median values of the appropriate DRL quantity observed at each health care facility.” Following ICRP convention, median values of CTDI_vol, dose-length product (DLP), and SSDE were calculated for each facility by size and by age bins. The 50th and 75th percentiles for these median values were then determined for each of the examinations.

Distributions of age for head and effective diameter for body examinations were obtained using univariable procedures, and these informed the age and size bins. The bins were constructed using the distribution of the data (ie, the number of data points in each of the bins, ranges of patient sizes, and whole-number bin sizes that facilities could use in clinical practice). All analyses were performed with SAS software (version 9.4, SAS Institute).

Results

The DIR collected data from 2 830 118 pediatric CT examinations from 2625 U.S. facilities between January 2016 and December 2020. Of these 2 830 118 examinations, 702 758 body examinations were missing effective diameter (25%) and 57 539 were multiphase examinations (2%). Both of these made up the largest exclusion criteria. The top 10 most frequently performed examinations, minus exclusions, yielded a total of 1 543 535 single-phase examinations from 1625 facilities (55% [1 543 535 of the 2 830 118 total pediatric CT examinations) to analyze (Tables 1, 2).

Head CT was the most commonly performed examination (876 655 of 1 543 535 examinations [57%]), followed by CT of the abdomen and pelvis (438 528 of 1 543 535 examinations [28%]), neck and cervical spine (129 347 of 1 543 535 examinations [8.4%]), and chest (72 228 examinations [4.7%]) (Table 1). Table 2 shows that 66% of 1625 participating facilities (1068 of 1625) were community hospitals, 16% (264 of 1625) were freestanding centers, 9.5% (154 of 1625) were academic facilities, and 3.5% (57 of 1625) of the facilities were dedicated children’s hospitals. Pediatric CT is performed mainly in community hospitals (874 501 of all 1 543 535 examinations analyzed [57%]) and not in dedicated children’s hospitals (407 288 of 1 543 535 examinations [26%]). Six hundred thirty of the 1625 facilities (39%) were in metropolitan areas, 595 facilities (37%) were in suburban areas, and 400 facilities (25%) were in rural areas. Two facilities (0.1%) contributing 1051 examinations did not indicate location but were included in the analysis. Almost 60% of examinations (908 204 of 1 543 535) are performed at facilities in metropolitan areas; only 14% of pediatric CT examinations (217 025 of 1 543 535) are performed at facilities in rural areas. Almost half (49%, 761 528 of 1 543 535 examinations) of the data analyzed were provided by facilities that performed fewer than 50 examinations per month. Facilities performing more than 100 examinations per month provided 33% (502 797 of 1 543 535) of all the examinations analyzed. Fewer than 50 pediatric CT examinations per month were performed at more than 90% (1468 of 1625) of facilities. More than 100 examinations per month were performed at 3.4% (56 of 1625) of the facilities. Fifty-two percent of the examinations (798 577 of 1 543 535) were of boys. The median age of patients was 14 years (boys, 13 years; girls, 15 years) with 54% (827 494 of 1 543 535 examinations) below age 15 years and 46% (716 041 of 1 543 535 examinations) in the 15–18 years age group.

Table 3 shows the variation of AD and DRL with age. The dose indexes increase with age as expected. The 6–18 years age group included 593 573 of 811 150 head examinations (73%), 25 606 of 27 840 sinus examinations (92%), and 33 743 of 37 561 maxillofacial examinations (90%). Similarly, for neck, cervical spine, chest, and abdomen and pelvis examinations, the maximum number of examinations were in the 15–18 years age group.

Table 4 shows the variation of AD and DRL for the body examinations as a function of the effective diameter. Most of the body examinations fell in the 24–28 cm effective diameter bin; these were 26% of the 34 104 chest examinations, 34% of the 214 613 abdomen and pelvis examinations, and 22% of the 12 190 chest, abdomen, and pelvis examinations. The median and 75th percentile CTDI_vol and SSDE for these examinations increased with patient size, especially with the very large sizes. The median DLP values also increased consistently from smaller to larger sizes.

Discussion

This study analyzed the 10 most frequently performed pediatric CT examinations, which comprised a total of 1 543 535 examinations representing 1625 practices participating in the American College of Radiology Dose Index Registry. We found that pediatric CT is performed mainly in community hospitals (874 501 of 1 543 535 examinations [57%]) as opposed to dedicated children’s hospitals (407 288 of 1 543 535 examinations [26%]), in metropolitan areas (908 204 of 1 543 535 examinations [59%]), and in practices that perform CT relatively infrequently. Head CT was the most commonly performed examination (876 655 of 1 543 535 examinations [57%]). Data available allowed for the determination of diagnostic reference levels (DRLs) for volume CT dose index, size-specific dose estimate, and dose-length product for each of these examinations across age and size bins, in turn providing opportunities for comparison of benchmarks with existing international DRLs.

To the best of our knowledge, the DIR data represent the largest systematic review configured according to the type of pediatric CT examinations. The DIR was launched in 2011 (10,11) and, as of March 2021, has collected data on 120 542 713 adult and pediatric examinations representing 2842 facilities. Regarding demographic information, our findings are consistent with those of Strauss et al (24) who found that most pediatric CT is performed at nonacademic facilities (79%) and not in dedicated children’s hospitals (24%). Our study also showed that 46% of pediatric patients undergoing CT (716 041 of 1 543 535 examinations) were within a small 3-year age range (15–18 years), which is consistent with other studies (14,16). This could be a result of older children beginning to show adult injury and disease patterns, a lower threshold for use of CT than nonionizing imaging pathways outside of children’s hospitals, preferential referral of young children to children’s hospitals for imaging evaluation, or relatively curtailed use of CT in younger children than in teenagers. For all examinations, the DRLs and ADs for CTDI_vol and DLP increased as a function of patient age, which indicated that facilities were adjusting CT protocols according to patient age or size, whether through manual techniques or automated dose modulation. The trend of dose indexes increasing with patient age or size is to be expected and consistent with other studies (14–16, 24). For size-based analyses of body examinations, this trend was also observed except in some bins for the very large patient sizes (36 to <40 cm or >40 cm). For these large patient sizes, the DRLs and ADs for CTDI_vol and DLP sometimes dropped slightly, which may be due to facilities adjusting their protocols for obese patients to reduce relatively high radiation exposure. Although this in turn may decrease image quality, the DIR does not collect image quality information, and image quality assessment is beyond the scope of this study. SSDE adjusts the phantom-based CTDI_vol for the size of the patient and gives a more representative estimate of patient dose.

National DRLs and ADs are developed from dose information collected from typical CT examinations performed using commonly practiced techniques that we must assume yield adequate image quality because the examinations were interpreted. Because they are based on the 75th and 50th percentiles of the facility medians, they do not provide guidance on minimum dose levels that could possibly be used to achieve adequate image quality. Rather, the DRL is used to identify situations where patient dose levels may be unusually high, and the AD serves as a goal for optimization efforts for techniques and technologies in widespread use, while maintaining adequate clinical image quality. Facilities should analyze and compare their median, size-grouped, or age-grouped dose indexes to the ADs and DRLs to determine if their dose indexes are unusually high or low (6). Although a large dose deviation between DRLs and an institution’s own protocol values can trigger a review, that does not provide enough information alone to optimize the protocol. Image quality must also be taken into consideration in evaluating dose profiles for CT protocols with each scanner to determine if protocols are optimized (24,25). The DRL approach may become a preferable, more realistic, and useful means for performance comparisons in accountability programs such as Leapfrog, which includes radiation dose information in its surveys of pediatric health care providers. Caution should be used when participating in such programs that assess facility dose performance by giving more favorable scores only if they fall below the medians of participating facilities. This method is flawed for many reasons, including the potential for promoting only a lower dose, potentially degrading image quality necessary for adequate diagnosis (26). DRLs and ADs are not intended to be used for comparisons to dose indexes for individual patients because the DRL process is intended for optimization of protection for groups of patients and is based on standard patients, not individual patients (6).

One of the advantages of using a registry to determine national ADs and DRLs is eliminating the need to manually collect data from a small sample of facilities and patients. Data from a very large patient population and an all-inclusive set of examinations is automatically collected resulting in fewer statistical sampling errors and enabling frequent updates. In addition to the recognized value in medical imaging, establishment of ADs and DRLs through the DIR may also serve as benchmarks for practice performance within the domain of the national DIR and also for comparison with existing international reports. Comparison of DRLs and ADs is most easily implemented by the facility if it has a system to automatically monitor patient dose indexes so that aggregate results may be evaluated.

It is important to note that the ICRP 135 method using facility median dose indexes to develop DRLs (6) gives equal weight to each facility irrespective of examination volume. This means that median dose indexes for examinations performed at the large number of community hospitals (n = 1068) relative to the small number of children’s hospitals (n = 57) more heavily weight the resultant DRLs and ADs, thus potentially resulting in higher DRLs than if dedicated pediatric imaging facilities were more heavily weighted or evaluated separately. Strauss et al (23) demonstrated that adult imaging facilities generally use higher radiation doses when imaging pediatric patients compared with dedicated pediatric facilities. However, the ICRP specifically intended that equal weighting across facilities would be used in the development of national DRLs (ie, “DRL values should be derived from a group of facilities that is both large enough and sufficiently diverse to represent the range of practices within the country or region [eg, the European Union] for the particular examination or procedure”) (6). National DRLs published for other countries do not break out their benchmarks by facility type (16,27–32). By calculating national DRLs using both pediatric and nonpediatric facilities, we propose these results as reasonable and achievable for pediatric and nonpediatric facilities alike following international guidelines.

Table 5 shows how the American College of Radiology DIR DRLs and ADs for CTDI_vol and DLP as a function of patient age compare with recent DRLs from other countries for three commonly reported examinations—CT of the head without contrast material, chest with contrast material, and abdomen and pelvis with contrast material. Although the current investigation separates these examinations into those with and those without contrast material, the international DRLs generally do not designate contrast material use for an examination. Although international pediatric DRL methods, age groupings, and results differ, thus making direct comparisons difficult, our results generally fall within the ranges reported across other countries (16,27–32).

Our study had several limitations. The registry is voluntary and is not a random sample of facilities, examinations, or patients for all U.S. CT practice. However, the DIR demographic information suggests broad participation from all types of facilities. Although 5 years of data were collected to ensure adequate sample size for each bin in each examination, it is possible that practice patterns may have changed during this period. Although combining with and without contrast material examinations for each body part would likely increase individual bin sample size over a shorter period of time, we felt it was important to use a similar method to that used in the initial adult report (9). This may change in future iterations of national DRLs. Another limitation is the current manual process for examination code mapping, with its inherent inconsistencies. Although the DIR drives facilities to standardize procedure names by means of mapping tools and RadLex^® terms (17), facilities may not tag examinations accurately, which may cause problems both by skewing the benchmark data and by being compared with inappropriate benchmark data itself. The DIR only collects information provided in the scanner’s radiation dose structured report, secondary capture dose screen, and Digital Imaging and Communications in Medicine, or DICOM, header and does not collect clinical information. The use of dose reduction techniques or clinical indication is unknown as these are not included in the DICOM header. Protocols, image quality, and thus dose vary based on the clinical indication. For example, the exposure can vary for a repeated assessment of ventricular size in a chronic condition of ventriculoperitoneal shunt presence versus level 1 trauma head CT although the RadLex^® Playbook identifier mapping for both would be the same. Indication-based DRLs for CT, such as those that have been investigated in Europe (33,34), would be of value. Indication-based DRLs would result in dose being adjusted according to clinical indication for the same anatomic region, and the concept of “one dose fits all” would not apply.

In conclusion, this work provides diagnostic reference levels (DRLs) and achievable doses (ADs) for the 10 most commonly performed pediatric CT examinations performed in the United States using the American College of Radiology CT Dose Index Registry (DIR) representing a broad representation of geography and practice types. This is also the first time that national pediatric DRLs and ADs have been developed as a function of both patient age and size. These results enable facilities to effectively compare their pediatric patient dose indexes to national benchmarks and to work to optimize their CT protocols resulting in an appropriate dose for diagnostic purposes. The DIR will update its ADs and DRLs on a routine basis to capture future trends in CT scanners, imaging protocols, and radiation dose.

Disclosures of Conflicts of Interest: K.M.K. No relevant relationships. P.F.B. No relevant relationships. M.B.C. No relevant relationships. J.W. No relevant relationships. E.S. No relevant relationships. M.S. No relevant relationships. D. Golden No relevant relationships. D. Gress No relevant relationships. J.B. No relevant relationships. W.S. Leadership or fiduciary role in American College of Radiology Dose Index Registry. K.S. No relevant relationships. D.F. Member of Radiology editorial board.

Author Contributions

Author contributions: Guarantors of integrity of entire study, K.M.K., P.F.B., M.B.C., D. Golden; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; agrees to ensure any questions related to the work are appropriately resolved, all authors; literature research, K.M.K., P.F.B., D. Golden, D. Gress, D.F.; clinical studies, K.M.K., J.W., E.S., D. Golden, D.F.; experimental studies, E.S., D. Golden; statistical analysis, P.F.B., M.B.C., E.S., M.S., D. Golden, D. Gress, W.S.; and manuscript editing, K.M.K., P.F.B., M.B.C., J.W., E.S., D. Golden, D. Gress, W.S., K.S., D.F.