Safely Shifting MRIs for Seizure Evaluation to the Outpatient Setting Free

Project members included pediatric hospitalists, a hospitalist quality-improvement expert, a pediatric neurologist, a physician utilization review manager, resident physicians, the pediatric department chair, and a financial analyst. This single-center initiative took place in a Mid-Atlantic suburban children’s hospital. The hospital has 226 pediatric beds and averages 7000 general pediatric admissions each year. There is 1 consulting pediatric neurology group with 9 neurologists who admit patients to the pediatric hospital medicine service. The population included patients aged <24 years discharged from general pediatric units with diagnostic codes of seizures, syncope, or movement disorders (Supplemental Table 3). Any patients requiring an ICU stay were excluded from data analysis, as were patients who were admitted for a scheduled MRI or EEG. Data were extracted from the electronic medical record (EMR; Epic Systems, Verona, Wisconsin) at the encounter level. The Revised Standards for Quality Improvement (QI) Reporting Excellence guidelines were used.¹⁰ 

To conduct this project, key stakeholders were gathered from the hospitalist division, emergency department, ICU, and neurology groups to analyze relevant literature^{3–5,11–20}  and develop criteria for MRI completion in patients presenting with seizure-like activity (Fig 1). This guideline established patient and seizure characteristics for which inpatient, outpatient, or no MRI were recommended. After implementation, the largest hurdle in shifting MRIs to the outpatient setting was the cumbersome process to obtain outpatient MRIs, especially those requiring anesthesia.

A detailed process map of the existing outpatient MRI scheduling process, including bottlenecks and delays, was created which highlighted areas for improvement (Fig 2). This process map also detailed the numerous team members working in coordination to obtain an outpatient brain MRI, including anesthesia, child life, radiology, and the patient’s primary pediatrician and caregiver. An ideal process map was created to automate and streamline processes, and reduce bottlenecks and caregiver burdens.

The interventions included addending the inpatient history and physical provider note with the components required for procedural clearance and creating a communication pathway with child life services to automate the pre-MRI anesthesia evaluation. The hospitalist team agreed to use EMR-based checklists, shortcuts, and smartphrases (blocks of text entered into documents using an abbreviation) to facilitate screening for history and physical addendums that meet preanesthesia requirements. Discharge instruction templates, including information for families regarding day-of-MRI instructions, and an outpatient MRI order set, including neurology contact information for the radiologist to communicate urgent results, were developed. To reduce the task burden on families, the following steps were either automated or removed: Scheduling and attending a neurology follow-up visit to obtain the outpatient MRI order, scheduling and attending a preanesthesia procedure visit with their primary physician for medical clearance, returning documentation of anesthesia clearance to the MRI schedulers, and calling to schedule the MRI procedure. In addition, a separate list was created for MRI schedulers in their EMR queue to quickly identify and prioritize scheduling for these patients.

Finally, education was provided to neurologists, hospitalists, and residents on the MRI decision guideline and new outpatient MRI ordering process. The Model for Improvement and Plan–Do–Study–Act cycles were used to test these interventions for meaningful changes in desired outcomes in real time and provide feedback to providers.

One aim of this project was to create a new MRI outpatient order set that would automatically cascade into the revised scheduling process. Initially, there was a desire for the consulting neurologists to order the outpatient MRI while the patient was still admitted; however, their support staff could not support the prior authorization process for patients who had not yet visited the neurology clinic. The scheduling office agreed to initiate the prior authorization process and the physician utilization review manager agreed to take on prior authorization denials that should arise from insurance companies. The consulting neurologist pager number was added to the outpatient MRI order set so that urgent results could be addressed in real time. Several months after the intervention was initiated, the outpatient order set was amended to automatically route the result to the patient’s primary care physician, in addition to the neurologist.

The pediatric order included preanesthesia instructions for a patient’s last meal and drink when needed. A template of discharge instructions for patients was devised, such that there would be planned redundancy between the order printout and the discharge instructions received by the family. Additionally, the EMR team helped create a separate screen within the radiology schedulers’ EMR to which these orders filtered, so they can be prioritized separately from the high volume of outpatient adult radiology orders.

Several months into the project, it was noted that some of the neurologists were recommending outpatient MRI if the patient presented with 2 generalized tonic-clonic seizures in 24 hours that were not prolonged, many of whom had an obvious viral trigger. After review with the neurology team, the algorithm was amended to allow patients with complex febrile seizures without concerning features to receive an outpatient MRI.

We analyzed data by encounter, rather than by patient, because our MRI guideline applied to first-time and subsequent seizure episodes, each of which may result in hospitalization. The primary outcome measure was the percentage of encounters that included an inpatient MRI. Secondary outcome measures were the percentage of patient encounters with seizure-like events who completed an outpatient MRI after hospital discharge and those who did not complete an MRI (neither inpatient nor outpatient), and the median hospital length of stay (LOS). Process measures were the percentage of encounters that followed the new MRI ordering guidelines and the new outpatient MRI order set. Balancing measures included the average time to completion of outpatient MRI and the percentage of outpatient MRI results that required intervention, including readmission.

Project members collected retrospective and prospective deidentified data from patient charts in the EMR for patients meeting inclusion criteria. Baseline data were from June 1, 2020, to June 30, 2021. For guideline adherence, we analyzed baseline data from November 1, 2020, to June 30, 2021, by reviewing patient and seizure characteristics and applying the guide retrospectively. Prospective data were collected and analyzed monthly from July 1, 2021, to December 31, 2022. Then, analysis was transitioned to a biweekly basis to identify and react to changes in a timelier manner. χ², Fisher’s exact, and Mann Whitney U testing were used to determine significant differences between groups.

Analysis was conducted in Excel’s QI Macros to produce statistical process control charts. Upper and lower control limits were defined as >3 SDs above and below the mean. Special cause variation was identified as 8 consecutive points above or below the average, or a single point outside of the upper or lower control limits.²¹  Patient encounters receiving an inpatient MRI were subanalyzed by admission diagnosis category, race, ethnicity, language, and payer, as well as by event type (Table 1).

This project was a QI activity and exempt from our local institutional review board.

The results from 410 pediatric hospital encounters for seizure-like events were analyzed, including 134 encounters before the intervention. There were 6 patients with >1 admission during the study period, with 2 hospital encounters per patient. There were no differences between preintervention and postintervention populations in terms of payer, language, or age (Table 2). However, in the postintervention population there were fewer white patients and more patients whose ethnicity data were unavailable. Although there were no differences between provoked and all other seizure types in the pre- and postintervention populations, the postintervention population had more patient encounters with provoked afebrile seizures and spells without abnormal movements, and fewer encounters with unprovoked seizures.

Encounters that included the completion of inpatient MRIs decreased from a mean of 50% to 26% (Fig 3a). The most significant reduction occurred for patients with complex febrile seizures (47%, n = 14 and 15%, n = 10, P value .0008), provoked but afebrile seizures (100%, n = 7 and 21%, n = 7, P value .0002), and unprovoked seizures (65%, n = 30 and 45%, n = 30, P value .039). Before the guideline implementation, 80% (n = 4) of Latinx patients received an inpatient MRI compared with 42% to 52% of other races, but this difference was not significant. There was a reduction in the variation by race from 38% to 10% after guideline implementation, but this was not significant. The percentage of encounters in which outpatient MRI was ordered increased from a mean of 3.5% to 14% (Fig 3b). The percentage of patient encounters which did not include an MRI (neither inpatient nor outpatient) remained at a mean of 46% (Fig 3c). The percentage of encounters for whom the MRI guideline recommendations were followed increased from a mean of 80% in the 7 months before the intervention to a mean of 93% in the final 12 months of the project (Fig 3d). Since its release, 75% (n = 24) of outpatient MRI orders used the new order set.

Eighty-six percent (n = 18) of outpatient MRIs ordered using the order set were completed (3 were canceled by the neurologist after discharge and removed from the denominator) compared with 60% (n = 9) of MRIs ordered without the order set. This difference was not significant, possibly because of the small number of outpatient MRI orders placed without using the order set (P value .12 via Fisher’s exact test). The time to outpatient MRI completion from the discharge date decreased from a median of 36.5 days to 30.0 days. However, this difference was not significant (P value .14 via Mann Whitney U testing).

There was no significant difference in abnormal MRIs among those completed during the hospitalization and as an outpatient (28.4%, n = 42 and 20.5%, n = 8, respectively, P value .57 via χ² testing). However, 3 patients who received inpatient MRIs required urgent intervention (2 for cerebral infarcts and 1 for elevated intracranial pressure); the MRI guideline was followed for all 3 encounters. None of the outpatient MRIs had abnormal results requiring readmission nor urgent intervention.

The average LOS per encounter remained unchanged (34 hours and 32 hours, P value .4 via Mann Whitney U testing).

We successfully reduced inpatient MRIs for patient encounters with seizure evaluations from 50% to 26%, exceeding our primary aim of a 10% reduction, without any outpatient MRI results requiring readmission or urgent intervention. The most significant reduction occurred in patients with complex febrile seizures, provoked afebrile seizures, and unprovoked seizures. We found that patients with a history of seizures or with spells less convincing for seizures did not have significant reductions in inpatient MRIs after MRI guideline implementation. We expected, but did not find, a reduction in LOS. This may be because of variation in length of EEG per consulting neurologist and is a focus for a future project.

In reviewing our results for inequity, there were no significant differences in patients who received inpatient MRIs by race, ethnicity, language, or payer before or after intervention implementation (all P values > .05 via χ² testing). As a point of discussion, 100% of uninsured patients (n = 5) received inpatient MRIs preintervention, whereas only 25% (n = 2) did postintervention. Those who did not receive an inpatient MRI did not require one (inpatient or outpatient), per the established MRI guideline.

Review of the literature reveals this project as novel. In a systematic review of interventions to reduce low-value imaging, only 18% were in pediatric patients and only 4% involved MRI studies.²²  Most pediatric studies involved reducing imaging in bronchiolitis and minor head injury. To our knowledge, this is the only published project decreasing inpatient MRI usage in pediatric patients with seizure-like events. The systematic review found that multicomponent interventions (guideline creation, stakeholder recruitment, provider education) are most effective.²²  Robinson et al developed a standardized workflow via e-mail that encouraged care teams to reconsider the utilization of inpatient brain MRI in an adult patient population to prevent discharge delays.²³  The study resulted in a decrease in inpatient MRI studies similar to our study by shifting these imaging studies to the outpatient setting or canceling them, and demonstrated cost savings. However, this study did not provide guidance on the safety of this shift. Additionally, they had a 59% response rate to their e-mailed requests to clinicians, which denoted poor buy-in. In our study, we improved buy-in through extensive stakeholder discussions on barriers and solutions to the outpatient MRI ordering process; guideline adherence was achieved in 93% of patients. In another study regarding effectiveness of clinical decision support in safe imaging reduction, Blackmore et al found a 23.2% reduction in rates of brain MRI studies for patients with headaches.²⁴  Our project also adds to existing pediatric studies that reduced head computed tomography scans in patients with minor head injury^25,26  and abdomen and pelvic computed tomography in patients with a suspicion of appendicitis.²⁷ 

The success of our project was dependent on several variables. The multidisciplinary approach was one of the key drivers of success. Engaging providers from the outset was critical to ensuring buy-in for the MRI guideline. The preintervention process map highlighted interdepartmental delays and the effort spent by the patient’s family to complete an outpatient MRI. By understanding the barriers through each stakeholder’s lens, collaborative solutions were created. Engaging representatives from anesthesia, radiology, and child life was key to creating a new process that minimized delays and barriers. By having open and constant communication between the representative from each department and the QI team, decisions and changes were made in real time. Finally, having leadership buy-in on the project was necessary given the institutional changes that came about as a result of the project. The process used for this project has the potential to be replicated by other departments and specialties to similarly transition inpatient management to the outpatient arena.

There are several limitations to this project. This initiative was conducted at a single institution so the results may not be generalizable to institutions with differing workflows. The presenting symptom of seizure-like activity is difficult to capture via diagnostic codes because it can range from abnormal movements to status epilepticus. Because of the broad range of clinical presentations of seizures, providers must be afforded flexibility with regard to the utilization of a guideline rather than an institutional policy.

Our results by race and ethnicity should be interpreted with caution because of a lack of data on race in 7% and 14% and on ethnicity in 6% and 17% of encounters in the pre- and postintervention populations, respectively. We were also unable to track outpatient MRIs that occurred outside of our EMR system at other institutions, which could have impacted our analysis of outpatient MRI completion. It should be noted that we were not able to directly calculate the cost savings of shifting MRIs to the outpatient setting; however, there were indirect cost-per-case savings from not pursuing an inpatient MRI for most patients. Finally, we were unable to assess family satisfaction with shifting MRIs to the outpatient arena, nor with the new outpatient scheduling process. This could be a future area of focus for this project.

This QI project achieved its aim of safely reducing inpatient MRI utilization from 50% to 26% within 6 months. The project’s process development is generalizable to other diagnoses and patient groups, and sustainable given the automated steps created. The next step of this project is to apply a similar process to EEG ordering practices.