Headache Neuroimaging

A Survey of Current Practice, Barriers, and Facilitators to Optimal Use

Evan L. Reynolds PhD; James F. Burke MD, MS; Lacey Evans MPH; Faiz I. Syed MD, MS; Eric Liao MD; Remy Lobo MD; Wade Cooper DO; Larry Charleston IV MD, MSc; Brian C. Callaghan MD, MS

Disclosures

Headache. 2022;62(1):36-56. 

In This Article

Results

Survey Respondent Characteristics

The survey was successfully completed by 431 of 1426 clinicians (30.2% response rate). Physicians accounted for 76.0% of responders and APCs accounted for 24.0% (nurse practitioners: 18.0% and physician assistants: 6.1%). Clinician specialties were neurology in 45.5%, primary care in 50.3% and other specialties in 4.2%. There were 183 (43.1%) clinicians that made headache neuroimaging decisions less than once per week, 92 (21.7%) once per week, 100 (23.5%) multiple times per week, 11 (2.6%) once per day, and 39 (9.2%) multiple times per day. The frequency of missing responses for individual survey items, stratified by clinician specialty and clinician type are reported in Table 1, Table 2, Table 3 and Table 4. Two clinicians did not report their specialty (neurology, primary care, or other) or type (physician or APC). Across individual survey items, nonresponse rate ranged from 0.01% to 7.9% with a mean of 5.4%.

Harms and Benefits of MRI

In patients with a normal neurologic examination, the median (25th percentile and 75th percentile) clinician reported finding any abnormality in 6%–10% (1%–5%, 11%–30%) of patients, an abnormality that caused changes in management in 1%–5% (<1%, 1%–5%) of patients and a brain tumor in less than <1% (<1% and <1%) of patients. APCs reported identifying brain tumors (APCs: <1% [<1%, 1%–5%] vs. physicians: <1% [<1%, <1%], p = 0.001) and abnormalities that resulted in changed patient management (APCs: 1%–5% [<1%, 1%–5%] vs. physicians: 1%–5% [<1%, 15–5%], p = 0.009) more often than physicians. Compared with primary care clinicians, neurologists more frequently reported identifying abnormalities (neurology: 6%–10% [1%–5%, 11%–30%] vs. primary care: 6%–10% [1%–5%, 11%–30%], p < 0.001), although management changes were inversely related (neurology: 1%–5% [<1%, 1%–5%] vs. primary care: <1% [<1%, 1%–5%], p = 0.027). Nevertheless, both primary care clinicians and neurologists reported similar rates of identifying a brain tumor (p = 0.196; Table 1).

Overall, more clinicians believe that patients with headaches benefit (occasionally/often/always: 62.5%) from decisions stemming from MRI results than are harmed (occasionally/often/always: 42.2%). Clinicians believed the most common harms stemming from false positive results included unnecessary consultations (often/always: 17.0%), tests (often/always: 11.9%), procedures (often/always: 5.8%), and medications (often/always: 4.0%). The definition of a false positive result was not specified in the survey but may have included both incidental and nonspecific findings. Neurologists reported harms less frequently than primary care clinicians (neurology: rarely [rarely, occasionally] vs. primary care: occasionally [rarely, occasionally], p = 0.004). Specifically, neurologists believed there were fewer unnecessary medications (neurology: rarely [rarely, rarely] vs. primary care: rarely [rarely, occasionally], p = 0.020) and consultations (neurology: occasionally [rarely, occasionally] vs. primary care: occasionally [rarely, occasionally], p = 0.002), but not tests (neurology: occasionally [rarely, occasionally] vs. primary care: occasionally [rarely, occasionally], p = 0.086) or procedures (neurology: rarely [rarely, occasionally] vs. primary care: rarely [rarely, occasionally], p = 0.228) following a false positive MRI finding. APCs believed patients with headaches were more likely to benefit from neuroimaging compared to physicians (APCs: occasionally [occasionally, often] vs. physicians: occasionally [rarely, occasionally], p < 0.001). Additionally, APCs less frequently reported patients having difficulty tolerating the MRI itself (APCs: occasionally [rarely, occasionally] vs. physicians: occasionally [rarely, occasionally], p = 0.008), and having unnecessary tests (APCs: rarely [rarely, occasionally] vs. physicians: occasionally [rarely, occasionally], p < 0.001), procedures (APCs: rarely [never, rarely] vs. physicians: rarely [rarely, occasionally], p < 0.001), and consultations (APCs: rarely [rarely, occasionally] vs. physicians: occasionally [rarely, occasionally], p < 0.001) due to false positive findings.

Overall, clinicians were more often emotionally burdened by harms from a guideline-supported decision to not perform (moderately/very much/extremely: 88.5%) than by harms from a decision to perform an MRI (moderately/very much/extremely: 77.6%). APCs were more often emotionally burdened when patients were harmed by performing (APC: very much [moderately, very much] vs. physicians: moderately [slightly, very much], p < 0.001) or not performing an MRI (APCs: very much [moderately, extremely] vs. physicians: very much [moderately, very much], p = 0.002) compared with physicians. There was no difference in the emotional burden between specialties following a decision stemming from an MRI.

Implementation Interventions

Interventions that most clinicians believed would very much or extremely improve MRI utilization included increased access to neurologist consultations (44.8%), less time pressure in the clinic (43.6%), and streamlined referral protocols (43.5%). The least popular interventions included restrictions on neuroimaging ordering within the electronic health record (EHR) (not at all/slightly: 54.7%), reminders about optimal practice patterns in the EHR (not at all/slightly: 49.3%) and patient education materials regarding the risks/benefits of headache MRI (not at all/slightly: 37.7%). APCs and primary care clinicians were consistently more receptive to the interventions compared with physicians and neurologists, respectively (p < 0.05 for 17/19 comparisons in Table 2).

Clinician Attitude and Confidence Towards MRI Utilization

Neurologists (good [fair, very good]) reported having significantly better knowledge of headache neuroimaging guidelines compared with primary care clinicians (fair [fair, good], p < 0.001) and physicians (good [fair, good]) reported having significantly better knowledge of guidelines compared to APCs (fair [fair, good], p < 0.001) (Table 3).

Most clinicians agree or strongly agree that, for patients with headaches, MRI is overused (73.9%) and that it is important to reduce MRI utilization (65.3%), but that it was their responsibility to never miss a brain tumor (80.4%). More physicians believed that MRIs are overused compared with APCs (physicians: agree [agree, strongly agree] vs. APCs: agree [neutral, agree], p < 0.001). Physicians were more likely to endorse that MRIs should be reduced in patients with headaches compared with APCs (physicians: agree [neutral, agree] vs. APCs: agree [neutral, agree], p = 0.007). There were no differing opinions of excess MRI utilization between neurologists and primary care clinicians.

Physicians had more confidence in their abilities to identify red flags (physicians: agree [agree, strongly agree] vs. APCs: agree [agree, agree], p < 0.001), identify abnormalities on a neurologic examination (physicians: agree [agree, strongly agree] vs. APCs: agree [agree, agree], p < 0.001), determine whether to order an MRI in patients with headaches (physicians: agree [agree, strongly agree] vs. APCs: agree [agree, agree], p < 0.001), discuss the decision not to order an MRI with patients with headaches (physicians: agree [agree, agree] vs. APCs: agree [agree, agree], p = 0.016), and interpret and appropriately act on MRI reports (physicians: agree [agree, strongly agree] vs. APCs: agree [neutral, agree], p < 0.001) compared with APCs. Neurologists had significantly more confidence in their abilities to identify red flags (neurology: strongly agree [agree, strongly agree] vs. primary care: agree [agree, agree], p < 0.001), identify abnormalities on a neurologic examination (neurology: strongly agree [agree, strongly agree] vs. primary care: agree [agree, agree], p < 0.001), determine whether to order an MRI in patients with headaches (neurology: agree [agree, strongly agree] vs. primary care: agree [agree, agree], p < 0.001), discuss the decision not to order an MRI with patients with headaches (neurology: agree [agree, strongly agree] vs. primary care: agree [neutral, agree], p < 0.001), and interpret and appropriately act on MRI reports (neurology: strongly agree [agree, strongly agree] vs. primary care: agree [agree, agree], p < 0.001), compared with primary care clinicians (Table 4).

Other Considerations for MRI Ordering: Patient, Time, and Other Concerns

Neurologists were more likely to order an MRI upon patient request (neurology: occasionally [rarely, occasionally] vs. primary care: occasionally [rarely, occasionally], p = 0.003), but there were no differences when the MRI was not clinically indicated (neurology: rarely [rarely, occasionally] vs. primary care: rarely [rarely, occasionally], p = 0.095).

Clinicians reported spending a median of 10–20 min to both order and not order an MRI for patients with headaches. Physicians spent less time ordering (physicians: 10–20 min [6–10 min, 10–20 min] vs. APC: 10–20 min [6–10 min, 20+ min], p = 0.023) and a similar amount of time not ordering (physicians: 6–10 min [6–10 min, 10–20 min] vs. APC: 10–20 min [6–10 min, 10–20 min], p = 0.055) MRIs for patients with headaches compared with APCs. Among specialties, there were no differences in time to order an MRI (neurology: 10–20 min [6–10 min, 10–20 min] vs. primary care: 10–20 min [6–10 min, 10–20 min], p = 0.969), however, neurologists were faster when not ordering an MRI (e.g., discussing the decision with patients and implementing alternative management strategies) compared with primary care clinicians (neurology: 6–10 min [6–10 min, 10–20 min] vs. primary care: 10–20 min [6–10 min, 10–20 min], p < 0.001).

Most clinicians never/rarely (95.3%) considered financial incentives when deciding whether to order an MRI for patients with headaches. Across clinician types and specialties, a similar proportion of clinicians often/always considered malpractice concerns when deciding whether to order an MRI (APCs: 29.1%, physicians: 25.9%, neurology: 20.5%, and primary care: 32.4%). Requirements prior to a specialist referral were often/always considered for 69.1% of non-neurologists (Table 4).

Scenarios of Patients With Headaches

In the low-risk scenario without red flags, on average, clinicians would order an MRI 30.9% (SD = 31.7 and missing = 7) of the time. In contrast, clinicians would order an MRI 91.7% (SD = 16.2 and missing = 8) of the time in the high-risk scenario (n = 420 and mean [SD) difference: 60.6%, SD = 33.8, and Wilcoxon signed rank test p value < 0.001). Clinicians also reported increased beliefs of finding abnormalities (21.2% [SD = 20.9 and missing = 6] vs. 70.4% [SD = 24.6 and missing = 11], n = 417, mean SD difference: 48.9% [SD = 29.1], Wilcoxon signed rank test p value < 0.001), and tumors (10.1% [SD = 19.9 and missing = 8] vs. 53.9% [SD = 29.2 and missing = 13], n = 413, mean SD difference: 43.9% [SD = 30.8], Wilcoxon signed rank test p value < 0.001) in the high-risk scenario compared with the low risk scenario without red flags.

Clinician beliefs in harms were similar (13.2% [SD = 15.5 and missing = 9] vs. 16.6% [SD = 17.6 and missing = 16], n = 411, mean SD difference: 3.3% [SD = 16.3], Wilcoxon signed rank test p value < 0.001] but belief in benefits (25.1% [SD = 28.7 and missing = 12] vs. 68.2% [SD = 27.3 and missing = 18], n = 408, mean SD difference: 43.1% [SD = 32.4], Wilcoxon signed rank test p value < 0.001) increased substantially in the high-risk scenario compared with a low risk scenario without red flags. In the low-risk scenario without red flags, APCs had higher utilization (38.6% [SD = 32.9 and missing = 2]) compared with neurologists (28.7% [SD = 32.8 and missing = 3]) and primary care physicians (26.4% [SD = 28.0 and missing = 2], Kruskal-Wallis test p value = 0.009).

In the presence of three potential red flags with limited evidence, clinician behavior changed substantially. MRI utilization increased by a mean (SD) of 36.3% (SD = 37.1, from 30.9% [SD = 31.7 and missing = 7] to 67.1% [SD = 31.9 and missing = 7], n = 421, Wilcoxon signed rank test p value < 0.001). This was accompanied by an increased belief in finding abnormalities (21.2% [SD = 20.9 and missing = 6] to 46.3% [SD = 28.3 and missing = 11], n = 417, mean SD difference: 24.8% [SD = 28.1], Wilcoxon signed rank test p value < 0.001), tumors (10.1% [SD = 19.9 and missing = 8] to 18.3% [SD = 23.3 and missing = 10], n = 418, mean SD difference: 8.1% [SD = 19.1], Wilcoxon signed rank test p value < 0.001) and perceived benefits (25.1% [SD = 28.7 and missing = 12] to 43.6% [SD = 31.1 and missing = 15], n = 411, mean SD difference: 18.6% [SD = 28.5], paired t-test p value < 0.001) compared to the low risk scenario without red flags.

Across scenarios, neurologists consistently had the smallest belief in harms from an MRI (low risk without red flags: neurology: 9.2% [SD = 10.5 and missing = 5], APCs: 15.5% [SD = 17.4 and missing = 0], primary care: 15.8% [SD = 17.6 and missing = 4], Kruskal-Wallis test p value < 0.001; low risk with three potential red flags: neurology: 12.3% [SD = 13.7 and missing = 6], APCs: 18.5% [SD = 18.1 and missing = 0], primary care: 15.8% [SD = 14.7 and missing = 4], Kruskal-Wallis test pvalue = 0.002; and high risk: neurology: 12.8% [SD = 14.2 and missing = 5], APCs: 21.7% [SD = 22.2 and missing = 2], primary care: 17.3% [SD = 16.3 and missing = 9], Kruskal-Wallis test p value = 0.002; Figure 1).

Figure 1.

Utilization, findings, and benefits of MRI in three headache neuroimaging scenarios. Mean percentage of time clinicians would order an MRI, find any abnormality (including false positives), find a brain tumor, and the percentage of time the MRI would result in any harm or benefit for three headache neuroimaging scenarios stratified by clinician type and specialty (APCs, neurologists, and primary care clinicians). APCs, advanced practice clinicians; MRI, magnetic resonance imaging

processing....