Reducing neuroimaging in seizures with LIMIT clinical decision instrument

Compared with unstructured physician discretion, using the Let’s Image Malignancy, Intracranial Hemorrhage, and Trauma (LIMIT) clinical decision instrument (CDI) for determining which patients should undergo brain computed tomography (CT), would have reduced CT usage by more than 13%. These were the findings published in the American Journal of Emergency Medicine.

Brain CT is often used by emergency physicians to assess intracranial pathology in patients with seizures. Between 2005 and 2013, the use of brain CT in the emergency department (ED) increased by 60%. However, the greater use of brain CT has not improved diagnosis. Moreover, CT scans do increase the risk of cancer, are expensive, and prolong ED visits. In an effort to reduce CT scans, emergency physicians have developed CDIs to help inform diagnostic decisions and improve efficiency.

The objective of the current study was to compare LIMIT CDI to unstructured clinical judgment in assessing the need for neuroimaging in patients with recurrent seizures.

This was a retrospective study of patients presenting with seizure between 2019 and 2020 at a blinded university hospital and blinded inner-city community hospital with approximately 85,000 and 45,000 annual emergency department visits, respectively. Adult patient (N=1108) medical records were assessed using the LIMIT CDIs by three researchers to determine whether the patient was eligible for brain CT and compared their findings with whether the patient was recommended for a brain CT during their hospital visit.

The patient population was aged mean 42 (range, 18-95) years, 45.7% were women, 50.4% were Black, 52.8% had medication non-compliance seizure, 21.5% breakthrough seizure, 12.5% had traumatic injury, 6.5% a history of intracranial hemorrhage, and 1.9% active malignancy.

Nine out of 10 patients were identified by the LIMIT CDI, corresponding with a 90% (95% CI, 54.1%-99.5%) sensitivity, 81.1% (95% CI, 78.6%-83.3%) specificity, 4.1% (95% CI, 2.0%-8.0%) positive predictive value, and 99.9% (95% CI, 99.3%-99.9%) negative predictive value.

All 10 patients were identified by unstructured physician discretion, corresponding with a 100% (95% CI, 65.5%-100.0%) sensitivity, 67.8% (95% CI, 64.9%-70.5%) specificity, 2.7% (95% CI, 1.4%-5.2%) positive predictive value, and 100.0% (95% CI, 99.4%-100.0%) negative predictive value.

The physicians ordered a total of 364 brain CT (33%). If the physicians would have used the LIMIT CDI, only 217 (20%) of scans would have been ordered, reducing the scan rate by 13.3%.

Stratified by clinical experience, attending physicians ordered the most scans (n=141) and physician assistants or nurse practitioners the fewest (n=5). Accuracy increased with experience, in which advanced practitioners were accurate 60% of the time and attending physicians only 2.1% of the time.

This study may have included some errors in patient classification as data were not collected in person.

This study found that the LIMIT CDI would have performed better than unstructured clinician judgment and reduced the amount of brain CT ordering by more than 13%.

“Although the LIMIT CDI needs to be validated in a larger set of patients, it performed better than unstructured physician judgement for evaluating need for emergent neuroimaging after recurrent seizures,” the researchers concluded. Neurology Advisor