Understanding Variation in Care: Guidelines, Value, and Equity

In this issue of Hospital Pediatrics, Shannon et al¹  conduct a retrospective cohort study evaluating trends and variation in the utilization of head computed tomography scans (CTs), brain MRIs, and head ultrasounds in NICUs within major children’s hospitals throughout the United States. From 2010 to 2021, there was a decrease in head CTs in the first half of the study period (2010–2015) before stabilizing in the latter half (2016–2021), concurrent with an increase in brain MRIs and head ultrasounds. Overall, there was significant variation in the utilization of head CTs between hospitals. On further examination of the 2021 cohort, the authors found that ventriculoperitoneal (VP) shunts, neoplasms, and seizures/encephalopathy were associated with the highest rate of head CTs. In addition, the widest variations in the use of head CTs between hospitals were in patients with seizures/encephalopathy, VP shunts, and infection.¹ 

The use of various sensitivity analyses to better understand the underlying reason for these findings is a major strength of this study. Although no single cause was identified, the authors discuss several plausible explanations, including institutional culture and the availability of other neuroimaging options. As the authors point out, head CTs expose infants to more ionizing radiation than MRIs and increase long-term risks of cancer. However, there are also other tradeoffs to consider; brain MRIs can cost up to twice as much as head CTs.²  Although sedation is often unnecessary for MRIs, infants must remain relatively still to obtain high-quality images. For larger and older infants in the NICU, sedation with opioids and benzodiazepines may cause short- and long-term adverse effects.³  In addition, MRIs are more time-consuming than head CTs, and many NICUs require additional personnel to accompany infants for the duration of the procedure.

The study describes an overall head CT rate of 3%, which we found to be high but reasonable, given the older NICU patient population. As the authors mention, the population included in the Pediatric Health Information System (PHIS) database is mostly made up of outborn patients who are referred to children’s hospitals, and, therefore, the findings may not be generalizable to community/perinatal center NICUs with inborn populations. Selection bias may exist because most of these outborn patients are likely to be “sicker” and in need of higher-level care. One could argue that critically ill patients may, therefore, need urgent and quick imaging modalities like head CTs. However, the diagnoses that had the widest variation in head CTs between hospitals (seizures/encephalopathy, VP shunts, infection) suggest opportunities for standardization of imaging modalities that limit ionizing radiation exposure. The Newborn Brain Society,⁴  an international multidisciplinary collaboration of clinicians, scientists, and parents dedicated to the advancement of newborn brain care, published guidelines recommending MRI as the preferred imaging modality for term neonates with encephalopathy.⁵  In addition, the American Academy of Pediatrics and American College of Obstetricians and Gynecologists published a joint task force report recommending MRI for neonatal encephalopathy and acknowledged that CTs are sensitive to hemorrhage but not brain injury in the first 24 to 48 hours.⁶  Moreover, one of the neonatology metrics for the US News & World Report rankings is the availability of a fast shunt MRI for hydrocephalus.⁷  Quality metrics can also be used to promote the standardization of care between children’s hospitals.

A large population of NICU patients who require brain imaging includes preterm infants, specifically very low birth weight infants. These infants are at higher risk of intraventricular hemorrhage and brain injury, and various imaging modalities are often used to screen for and diagnose these injuries. To reduce variation in practice, the American Academy of Pediatrics recently published guidelines regarding routine neuroimaging of preterm infants. Interestingly, one of their recommendations is that head CT and MRI are not indicated as routine procedures in identifying neonates with brain injury.⁸  This is further supported by the Choosing Wisely list for newborn medicine, an effort to reduce low-value tests and treatments, which recommends avoiding routine screening of preterm infants with term-equivalent MRIs because there is insufficient evidence that such imaging will improve long-term outcomes.⁹  On the other hand, advocates of routine MRIs for preterm infants argue that the detection of brain injury or altered brain development may inform families and facilitate access to early intervention services, which may be a limited resource in some states.¹⁰  Nonetheless, guidelines such as the ones described above may be helpful to reduce variation and improve outcomes for infants with seizures, VP shunts, and infection.

Interestingly, the authors do not stratify by race or insurance status in their analysis as many previous PHIS studies have done.^11–13  Some hospitals may lean toward head CTs because of the limited availability of MRIs, thereby reflecting structural inequities in health care resources. In contrast, NICUs with MRIs in closer proximity, or even within the unit,¹⁴  may default to ordering MRIs because of convenience. More recently, PHIS has incorporated the Child Opportunity Index (COI), which is a publicly available surveillance tool that assesses attributes of neighborhood conditions that may promote or inhibit healthy child development, into its database. The COI includes 29 indicators across 3 domains: education, health and environment, and social and economic.¹⁵  The COI has also been used in several PHIS studies as a predictor of readmission for diabetic ketoacidosis,¹⁶  emergency department revisits,¹⁷  acute care use in children with medical complexity,¹⁸  and access to pediatric surgical care.¹⁹  Social determinants of health highlight further disparities that may impact care and can inform policymakers regarding the allocation of resources.

Given that the rate of head CTs ranged from 0% to 6.2% after risk adjustment, it may be helpful to focus on the outliers to understand the source of variation. Ideally, an evaluation of any difference in efficacy or safety outcomes between the outliers would also facilitate a value comparison between head CTs versus other imaging modalities.²⁰  We can only speculate what the “right” rate for head CTs should be for infants admitted to NICUs within children’s hospitals, but the number is likely not 0. As shown by Shannon et al, nevertheless, there remains room for improvement to reduce variation in care and identify high-value interventions that can be equitably distributed.