Schwedt, Todd J

Abstract Objective To develop a natural language processing (NLP) algorithm that can accurately extract headache frequency from free‐text clinical notes. Background Headache frequency, defined as the number of days with any headache in a month (or 4 weeks), remains a key parameter in the evaluation of treatment response to migraine preventive medications. However, due to the variations and inconsistencies in documentation by clinicians, significant challenges exist to accurately extract headache frequency from the electronic health record (EHR) by traditional NLP algorithms. Methods This was a retrospective cross‐sectional study with patients identified from two tertiary headache referral centers, Mayo Clinic Arizona and Mayo Clinic Rochester. All neurology consultation notes written by 15 specialized clinicians (11 headache specialists and 4 nurse practitioners) between 2012 and 2022 were extracted and 1915 notes were used for model fine‐tuning (90%) and testing (10%). We employed four different NLP frameworks: (1) ClinicalBERT (Bidirectional Encoder Representations from Transformers) regression model, (2) Generative Pre‐Trained Transformer‐2 (GPT‐2) Question Answering (QA) model zero‐shot, (3) GPT‐2 QA model few‐shot training fine‐tuned on clinical notes, and (4) GPT‐2 generative model few‐shot training fine‐tuned on clinical notes to generate the answer by considering the context of included text. Results The mean (standard deviation) headache frequency of our training and testing datasets were 13.4 (10.9) and 14.4 (11.2), respectively. The GPT‐2 generative model was the best‐performing model with an accuracy of 0.92 (0.91, 0.93, 95% confidence interval [CI]) and R 2 score of 0.89 (0.87, 0.90, 95% CI), and all GPT‐2–based models outperformed the ClinicalBERT model in terms of exact matching accuracy. Although the ClinicalBERT regression model had the lowest accuracy of 0.27 (0.26, 0.28), it demonstrated a high R 2 score of 0.88 (0.85, 0.89), suggesting the ClinicalBERT model can reasonably predict the headache frequency within a range of ≤ ± 3 days, and the R 2 score was higher than the GPT‐2 QA zero‐shot model or GPT‐2 QA model few‐shot training fine‐tuned model. Conclusion We developed a robust information extraction model based on a state‐of‐the‐art large language model, a GPT‐2 generative model that can extract headache frequency from EHR free‐text clinical notes with high accuracy and R 2 score. It overcame several challenges related to different ways clinicians document headache frequency that were not easily achieved by traditional NLP models. We also showed that GPT‐2–based frameworks outperformed ClinicalBERT in terms of accuracy in extracting headache frequency from clinical notes. To facilitate research in the field, we released the GPT‐2 generative model and inference code with open‐source license of community use in GitHub. Additional fine‐tuning of the algorithm might be required when applied to different health‐care systems for various clinical use cases.

Abstract Objective To compare the artificial intelligence‐enabled electrocardiogram (AI‐ECG) atrial fibrillation (AF) prediction model output in patients with migraine with aura (MwA) and migraine without aura (MwoA). Background MwA is associated with an approximately twofold risk of ischemic stroke. Longitudinal cohort studies showed that patients with MwA have a higher incidence of developing AF compared to those with MwoA. The Mayo Clinic Cardiology team developed an AI‐ECG algorithm that calculates the probability of concurrent paroxysmal or impending AF in ECGs with normal sinus rhythm (NSR). Methods Adult patients with an MwA or MwoA diagnosis and at least one NSR ECG within the past 20 years at Mayo Clinic were identified. Patients with an ECG‐confirmed diagnosis of AF were excluded. For each patient, the ECG with the highest AF prediction model output was used as the index ECG. Comparisons between MwA and MwoA were conducted in the overall group (including men and women of all ages), women only, and men only in each age range (18 to <35, 35 to <55, 55 to <75, ≥75 years), and adjusted for age, sex, and six common vascular comorbidities that increase risk for AF. Results The final analysis of our cross‐sectional study included 40,002 patients (17,840 with MwA, 22,162 with MwoA). The mean (SD) age at the index ECG was 48.2 (16.0) years for MwA and 45.9 (15.0) years for MwoA ( p < 0.001). The AF prediction model output was significantly higher in the MwA group compared to MwoA (mean [SD] 7.3% [15.0%] vs. 5.6% [12.4%], mean difference [95% CI] 1.7% [1.5%, 2.0%], p < 0.001). After adjusting for vascular comorbidities, the difference between MwA and MwoA remained significant in the overall group (least square means of difference [95% CI] 0.7% [0.4%, 0.9%], p < 0.001), 18 to <35 (0.4% [0.1%, 0.7%], p = 0.022), and 35 to <55 (0.5% [0.2%, 0.8%], p < 0.001), women of all ages (0.6% [0.3%, 0.8%], p < 0.001), men of all ages (1.0% [0.4%, 1.6%], p = 0.002), women 35 to <55 (0.6% [0.3%, 0.9%], p < 0.001), and men 18 to <35 (1.2% [0.3%, 2.1%], p = 0.008). Conclusions Utilizing a novel AI‐ECG algorithm on a large group of patients, we demonstrated that patients with MwA have a significantly higher AF prediction model output, implying a higher probability of concurrent paroxysmal or impending AF, compared to MwoA in both women and men. Our results suggest that MwA is an independent risk factor for AF, especially in patients <55 years old, and that AF‐mediated cardioembolism may play a role in the migraine–stroke association for some patients.

Abstract Background and Objectives Prior studies have established an association between a history of abuse and the development of migraine. This cross‐sectional observational study explored the relationship between self‐reported abuse history with migraine‐related sensory hypersensitivity symptoms. Methods In total, 588 adult patients with migraine from the American Registry for Migraine Research completed questionnaires: Generalized Anxiety Disorder‐7, Patient Health Questionnaire‐2 for depression, Photosensitivity Assessment Questionnaire, Hyperacusis Questionnaire, and Allodynia Symptom Checklist. Using four binary screening questions, patients were asked to self‐report if they believed they had suffered emotional, physical, or sexual abuse in their lifetime. Differences in questionnaire scores between groups with and without a history of abuse were determined. Regression models adjusted for age, sex, and basic headache features analyzed the relationship between abuse history and sensory hypersensitivity symptoms. Moderation analysis explored the role of headache frequency in this relationship. Mediation analysis assessed the indirect (Mediated) effect (IE) of abuse on sensory hypersensitivity through depression or anxiety. Additional models analyzed relationships between sensory hypersensitivity symptoms and abuse subtypes or the number of abuse subtypes. Results Of 588 participants, 222 (38%) reported a history of abuse. Patients with a history of abuse reported statistically significantly greater average headache frequency (7.6 vs. 4.7 days, p = 0.030). Patients with a history of abuse also reported higher average or median questionnaire scores: anxiety (7.6 vs. 4.7, p < 0.001, d = 0.56), depression (1.7 vs. 1.3, p = 0.009, d = 0.24), photophobia (0.54 vs. 0.44, p < 0.001, d = 0.32), hyperacusis (19.6 vs. 14.9, p < 0.001, d = 0.49), ictal allodynia (6.0 vs. 3.0, p < 0.001, d = 0.46), and interictal allodynia (1.0 vs. 0.0, p < 0.001, d = 0.30). After controlling for patient age, sex and years lived with headache, abuse maintained a significant association with every sensory hypersensitivity measure. Headache frequency significantly moderated the relationship between a history of abuse with increased ictal allodynia ( p = 0.036). Anxiety significantly mediated the relationships between abuse with photophobia (IE = 0.03, 95% CI = 0.01–0.04), hyperacusis (IE = 1.51, 95% CI = 0.91–2.24), ictal allodynia (IE = 0.02, 95% CI = 0.01–0.04), and interictal allodynia (IE = 0.02, 95% CI = 0.01–0.06). Depression significantly mediated the relationship between abuse with photophobia (IE = 0.02, 95% CI = 0.01–0.03) and with hyperacusis (IE = 0.45, 95% CI = 0.11–0.88). The association between the individual subtypes of abuse and the number of subtypes of abuse with sensory hypersensitivity symptoms varied. Conclusion A history of abuse is associated with greater migraine‐related sensory hypersensitivity symptoms. To reduce the impact of abuse on migraine symptoms, future studies should explore mechanistic connections between abuse and migraine‐associated symptoms.

Background and Objectives The association between migraine, depression, and anxiety has been established, but the impact of these psychiatric comorbidities on functional impairment in people with migraine has been under‐investigated. The purpose of this cross‐sectional observational study was to investigate the relationship between anxiety and depression symptoms on migraine‐related disability, pain interference, work interference, and career success in a cohort of patients with migraine. Methods This analysis included 567 migraine patients who had been enrolled into the American Registry for Migraine Research (ARMR) between February 2016 and June 2019. Patients completed the Generalized Anxiety Disorder‐7 (GAD‐7) and Patient Health Questionnaire‐2 (PHQ‐2) to measure symptoms of anxiety and depression, respectively. Patients completed the Migraine Disability Assessment Scale (MIDAS), Pain Interference (PROMIS pain short) questionnaire, and Work Productivity and Activity Interference (WPAI) questionnaire to measure levels of functional impairment at work and in daily activities. In addition, patients answered questions designed for ARMR regarding education and career interference. Models were created to describe the relationship between severity of psychiatric symptoms (anxiety and depression), and each outcome of interest (WPAI, MIDAS, pain interference, reporting that migraine had interfered with career success). Each model was controlled for age, sex, headache frequency, years with migraine, and average headache intensity. Results Among the 567 patients with migraine, mean (SD) age was 47.1 (13.7), 87.3% were female, and average headache frequency was 19.1 (9.3) days/month. PHQ‐2 scores were positively associated with scores on MIDAS ( b = 0.06, SE = 0.01, P ≤ .001), pain interference ( b = 1.4, SE = 0.2, P < .001), and WPAI including absenteeism ( b = 0.16, SE = 0.04, P = .007), presenteeism ( b = 2.7, SE = 1.1, P = .012), overall work productivity impairment ( b = 3.7, SE = 1.2, P = .001), and activity impairment ( b = 3.0, SE = 1.2, P = .009). PHQ‐2 scores were also associated with reporting that migraine interfered with career success ( b = 0.34, SE = 0.08, P ≤ .001). GAD‐7 scores were not associated with MIDAS, pain interference, WPAI, or reduced career success. Conclusions Severity of depression symptoms in patients with migraine is associated with migraine‐related disability, work interference, pain interference, and reduced career success. Patients with more severe symptoms of depression are more likely to have greater functional impairment. A management approach that addresses depression in those with migraine may lead to improvements in patient functioning.

Objective To summarize the current literature on non‐steroidal anti‐inflammatory drug and corticosteroid use during the coronavirus disease 2019 (COVID‐19) pandemic, recognizing that these are commonly used treatments in the field of headache medicine. Background The use of non‐steroidal anti‐inflammatory drugs and corticosteroids in patients during the COVID‐19 pandemic has been a controversial topic within the medical community and international and national health organizations. Lay press and social media outlets have circulated opinions on this topic despite the fact that the evidence for or against the use of these medications is sparse. In the field of headache medicine, these medications are used commonly and both patients and clinicians may have questions or hesitations pertaining to their use during the COVID‐19 pandemic. Methods A detailed search of the scientific and popular literature was performed. Results There is limited literature pertaining to the safety of non‐steroidal anti‐inflammatory drugs and corticosteroids during the COVID‐19 pandemic. To date, there are no clear scientific data that preclude the use of non‐steroidal anti‐inflammatory drugs in the general population who may acquire COVID‐19 or in those acutely infected with the virus. Several health organizations have concluded that treatment with corticosteroids during active infection should be avoided due to concerns of prolonged viral shedding in the respiratory tract and the lack of survival benefit based on the data from past coronaviruses and influenza virus; specific exceptions exist including treatment for underlying asthma or chronic obstructive pulmonary disease, septic shock, and acute respiratory distress syndrome. Conclusion Scientific information regarding the COVID‐19 pandemic is constantly evolving, and limited or contradictory information can lead to confusion for both patients and clinicians. It is recommended that prior to prescribing non‐steroidal anti‐inflammatory drugs and steroids for the treatment of headache, clinicians have open discussions with their patients about the potential risks and benefits of using these medications during the COVID‐19 pandemic. This manuscript summarizes the currently available evidence and understanding about these risks and benefits to help clinicians navigate such discussions.

Abstract Background Close associations between insomnia with other clinical factors have been identified in migraine, but there have been few studies investigating associations between insomnia and clinical factors in patients with persistent post-traumatic headache (PPTH). The study objective was to contrast the severity of insomnia symptoms in PPTH, migraine, and healthy controls, and to identify factors associated with insomnia in patients with PPTH vs. migraine. Methods In this cross-sectional cohort study, 57 individuals with PPTH attributed to mild traumatic brain injury, 39 with migraine, and 39 healthy controls were included. Participants completed a detailed headache characteristics questionnaire, the Migraine Disability Assessment Scale (MIDAS), Insomnia Severity Index (ISI), Hyperacusis Questionnaire (HQ), Allodynia Symptom Checklist, Photosensitivity Assessment Questionnaire, Beck Depression Inventory (BDI), State-Trait Anxiety Inventory, Post-Traumatic Stress Disorder (PTSD) checklist, Ray Auditory Verbal Learning Test, and the Trail Making Test A and B to assess headache characteristics, disability, insomnia symptoms, sensory hypersensitivities, and neuropsychological factors. Fisher’s test and one-way ANOVA or Tukey’s Honest Significant Difference were used to assess group differences of categorical and continuous data. Stepwise linear regression analyses were conducted to identify clinical variables associated with insomnia symptoms. Results Those with PPTH had significantly higher ISI scores (16.7 ± 6.6) compared to migraine patients (11.3 ± 6.4) and healthy controls (4.1 ± 4.8) ( p < 0.001). For those with PPTH, insomnia severity was most strongly correlated with the BDI (Spearman's rho ( Common.EditSubmissionSteps.Transform.EquationText ) = 0.634, p < 0.01), followed by Trait Anxiety ( Common.EditSubmissionSteps.Transform.EquationText = 0.522, p < 0.01), PTSD ( Common.EditSubmissionSteps.Transform.EquationText = 0.505, p < 0.01), HQ ( Common.EditSubmissionSteps.Transform.EquationText = 0.469, p < 0.01), State Anxiety ( Common.EditSubmissionSteps.Transform.EquationText = 0.437, p < 0.01), and MIDAS scores ( Common.EditSubmissionSteps.Transform.EquationText = 0.364, p < 0.01). According to linear regression models, BDI, headache intensity, and hyperacusis scores were significantly positively associated with insomnia severity in those with PPTH, while only delayed memory recall was negatively associated with insomnia severity in those with migraine. Conclusions Insomnia symptoms were more severe in those with PPTH compared to migraine and healthy control cohorts. Depression, headache intensity, and hyperacusis were associated with insomnia in individuals with PPTH. Future studies should determine the bidirectional impact of treating insomnia and its associated symptoms.

Abstract Background: Close associations between insomnia with other clinical factors have been identified in migraine, but there have been few studies investigating associations between insomnia and clinical factors in patients with persistent post-traumatic headache (PPTH). The study objective was to contrast the severity of insomnia symptoms in PPTH, migraine, and healthy controls, and to identify factors associated with insomnia in patients with PPTH vs. migraine. Methods: In this cross-sectional cohort study, 57 individuals with PPTH attributed to mild traumatic brain injury, 39 with migraine, and 39 healthy controls were included. Participants completed a detailed headache characteristics questionnaire, the Migraine Disability Assessment Scale (MIDAS), Insomnia Severity Index (ISI), Hyperacusis Questionnaire (HQ), Allodynia Symptom Checklist, Photosensitivity Assessment Questionnaire, Beck Depression Inventory (BDI), State-Trait Anxiety Inventory, Post-Traumatic Stress Disorder (PTSD) checklist, Ray Auditory Verbal Learning Test, and the Trail Making Test A and B to assess headache characteristics, disability, insomnia symptoms, sensory hypersensitivities, and neuropsychological factors. Fisher’s test and one-way ANOVA were used to assess group differences of categorical and continuous data. Stepwise logistic regression analyses were conducted to identify clinical variables associated with insomnia symptoms. Results: Those with PPTH had significantly higher ISI scores (16.72 ± 6.58) compared to migraine patients (11.28 ± 6.42) and healthy controls (4.05 ± 4.80) (p < 0.001). For those with PPTH, insomnia severity was most strongly correlated with the BDI (Spearman's rho (ρ) = 0.634, p < 0.01), followed by Trait Anxiety (ρ = 0.522, p < 0.01), PTSD (ρ = 0.505, p < 0.01), HQ (ρ = 0.469, p < 0.01), State Anxiety (ρ = 0.437, p < 0.01), and MIDAS scores (ρ = 0.364, p < 0.01). According to linear regression models, BDI, headache intensity, and hyperacusis scores were significantly positively associated with insomnia severity in those with PPTH, while only delayed memory recall was negatively associated with insomnia severity in those with migraine. Conclusions: Insomnia symptoms were more severe in those with PPTH compared to migraine and healthy control cohorts. Depression, headache intensity, and hyperacusis were associated with insomnia in individuals with PPTH. Future studies should determine the bidirectional impact of treating insomnia and its associated symptoms.

Objective The objective of this study was to identify brain regions having aberrant pain-induced activation in migraineurs, thereby gaining insight into particular aspects of pain processing that are atypical in migraineurs. Methods Functional magnetic resonance imaging assessed whole brain responses to painful heat in 24 adult episodic migraineurs who were at least 48 hours pain free and 27 healthy controls. Regions differentially activated in migraineurs compared to controls were identified. Activation intensities in these regions were correlated with headache frequency, number of migraine years, and time to next migraine attack. Results Migraineurs had greater pain-induced activation of lentiform nucleus, fusiform gyrus, subthalamic nucleus, hippocampus, middle cingulate cortex, premotor cortex, somatosensory cortex, and dorsolateral prefrontal cortex, and less activation in precentral gyrus and superior temporal gyrus. There were significant correlations between activation strength and headache frequency for middle cingulate ( r = 0.627, p = 0.001), right dorsolateral prefrontal cortex ( r = 0.568, p = 0.004), left fusiform gyrus ( r = 0.487, p = 0.016), left precentral gyrus ( r = 0.415, p = 0.044), and left hippocampus ( r = 0.404, p = 0.050) and with number of migraine years for left fusiform gyrus ( r = 0.425, p = 0.038). There were no significant correlations between activation strength and time to next migraine attack. Conclusions The majority of regions with enhanced pain-induced activation in headache-free migraineurs participate in cognitive aspects of pain perception such as attending to pain and pain memory. Enhanced cognitive pain processing by migraineurs might reflect cerebral hypersensitivity related to high expectations and hypervigilance for pain.