Research Article: Impact of early headache neuroimaging on time to malignant brain tumor diagnosis: A retrospective cohort study

Date Published: February 1, 2019

Publisher: Public Library of Science

Author(s): Matthew R. Carey, Brian C. Callaghan, Kevin A. Kerber, Lesli E. Skolarus, James F. Burke, Jonathan H. Sherman.

http://doi.org/10.1371/journal.pone.0211599

Abstract

Neuroimaging for headaches is both common and costly. While the costs are well quantified, little is known about the benefit in terms of diagnosing pathology. Our objective was to determine the role of early neuroimaging in the identification of malignant brain tumors in individuals presenting to healthcare providers with headaches.

This was a retrospective cohort study using administrative claims data (2001–2014) from a US insurer. Individuals were included if they had an outpatient visit for headaches and excluded for prior headache visits, other neurologic conditions, neuroimaging within the previous year, and cancer. The exposure was early neuroimaging, defined as neuroimaging within 30 days of the first headache visit. A propensity score-matched group that did not undergo early neuroimaging was then created. The primary outcome was frequency of malignant brain tumor diagnoses and median time to diagnosis within the first year after the incident headache visit. The secondary outcome was frequency of incidental findings.

22.2% of 180,623 individuals had early neuroimaging. In the following year, malignant brain tumors were found in 0.28% (0.23–0.34%) of the early neuroimaging group and 0.04% (0.02–0.06%) of the referent group (P<0.001). Median time to diagnosis in the early neuroimaging group was 8 (3–19) days versus 72 (39–189) days for the referent group (P<0.001). Likely incidental findings were discovered in 3.17% (3.00–3.34%) of the early neuroimaging group and 0.66% (0.58–0.74%) of the referent group (P<0.001). Malignant brain tumors in individuals presenting with an incident headache diagnosis are rare and early neuroimaging leads to a small reduction in the time to diagnosis.

Partial Text

Headache diagnoses are common, resulting in numerous outpatient visits, emergency department encounters, prescriptions, and diagnostic tests [1–5]. From 2007 to 2010, there were 51.1 million headache visits in the United States, resulting in 6.3 million neuroimaging studies that cost $3.9 billion [3]. However, the potential downsides of overly broad neuroimaging are numerous: unnecessary costs, inconvenience for the individuals, and identification of incidental findings that may result in downstream harms [6–8]. Therefore, the appropriateness of neuroimaging in many individuals with headaches has been called into question [9, 10]. The American Board of Internal Medicine Foundation Choosing Wisely campaign has identified neuroimaging for headaches as an area for more judicious use [11, 12].

We performed a retrospective cohort study using de-identified administrative claims data from the Clinformatics Datamart (OptumInsight: Eden Prairie, MN), a comprehensive database for all enrollees at a large, national insurer for years 2001 to 2014 [20]. The use of a de-identified database was determined to be exempt by the University of Michigan Institutional Review Board. An online appendix provides additional methodologic information (see S1 File).

In this national cohort study exploring early headache neuroimaging, we found that the rate of malignant brain tumors among patients presenting for the first time with headaches is low: 303 such individuals must be imaged to diagnose a single malignancy. Furthermore, our data suggest that current headache neuroimaging practices rarely result in missed or delayed malignant brain tumor diagnoses. Finally, neuroimaging is much more likely to identify probable incidental findings than malignancies. Providers should consider these data in the context of the full clinical picture before referring patients for neuroimaging.

 

Source:

http://doi.org/10.1371/journal.pone.0211599

 

Leave a Reply

Your email address will not be published.