Building Better Algorithms for the Diagnosis of Nontraumatic Headache

Nontraumatic headaches are challenging for general practitioners (GPs) and emergency department (ED) physicians to diagnose. Studies have shown that migraine headaches in these settings are often misdiagnosed and that patients do not receive proper treatment.¹ For this reason, researchers are creating algorithms for the differential diagnosis of headache; however, the screening tools for migraine assessment vary among these algorithms, and researchers debate their effectiveness.

"Up until now, there has been a lack of simple and accurate screening tools for migraine, which has led to difficulty in the differential diagnosis of the patient presenting with headache," said Christopher Booth, MD, a medical oncologist at Queens University in Kingston, Canada. His research team recently created an algorithm based on variables for the identification of migraine developed by Michel and colleagues.²

Studies have shown that fewer than half of patients with migraine are given the correct diagnosis, and only one third of these patients receive migraine-specific treatment.^3,4 GPs and ED physicians often use narcotics or a non-narcotic analgesic (eg, ketorolac) along with treatment for nausea, as opposed to using an antinausea and a migraine-specific medication (eg, a triptan or dihydroergotamine), said Deborah Isa Friedman, MD, professor in the departments of ophthalmology and neurology at the University of Rochester, New York. She is working to create an algorithm for the differential diagnosis of nontraumatic headache in the ED based on Lipton's ID Migraine scale.⁵

"A number of published studies have shown that diagnosis of migraine in the ED is abysmal, but physicians in the ED are still reluctant to label patients as having migraines," said Friedman. According to Friedman, migraine and other forms of nontraumatic headache account for roughly 3% of all ED visits. A small percentage of patients with migraine make up the vast majority of headache visits to the ED, and many of these patients with migraine are seeking first-time treatment. In an interview with Applied Neurology, Friedman cited a study by Blumenthal and colleagues¹ in which it was reported that 30% of patients who had been treated for headache in the ED still had the same symptoms 24 hours after discharge.

Friedman went on to say that the lack of proper diagnosis could also lead to unnecessary diagnostic imaging. "A lot of people with migraine get imaged every time they come to the ED because nobody wants to miss anything serious, even when a patient with known migraine reports that they are having their usual headache," she said. "There are a plethora of unnecessary scans that accumulate on these people."

That's why it is so important for physicians to develop algorithms for the differential diagnosis of headache in these settings, Friedman said. "There's this balancing act between identifying migraine and treating it properly but not missing anything."

"POUNDing" ALGORITHM

To create an algorithm to diagnose nontraumatic headache, Booth and colleagues reviewed articles that assessed the performance of various models developed for the diagnosis of migraine.⁶ The group found that many of these models had overlapping recommendations for the diagnostic criteria of headaches. They determined that the most effective model had been created by Michel and colleagues.² That model includes the following evaluation criteria (Figure 1): pulsatile quality, duration of 4 to 72 hours, unilateral location, nausea or vomiting, and disabling intensity. The team dubbed this the POUNDing scale, and concluded that any 4 of the 5 criteria strongly support a diagnosis of migraine headache.

If a diagnosis of migraine is likely, therapy should be initiated. Patients in whom there is any diagnostic uncertainty should be referred to a neurologist and/or be considered for neuroimaging.

Also included in the algorithm by Booth and colleagues is a component to determine whether a patient with headache should undergo neuroimaging. The algorithm states that patients with new headache onset, recently changed headache, or adult onset migraine should be considered for neuroimaging. Patients with thunderclap headache should be sent for neuroimaging and lumbar puncture and may also be referred to a neurologist. Patients with chronic headache should be evaluated for additional high-risk features during history taking and physical examination. If features are present, the patient should be sent for neuroimaging.

"We hope that the simple and accurate POUNDing mnemonic as described in our paper will aid clinicians in making a diagnosis of migraine," said Booth. He does not know of any clinics or EDs that have formally adopted the algorithm.

FRIEDMAN'S ALGORITHM

Friedman said she created her algorithm using evidence from the literature and clinical experience to make decision points that could easily be followed. The results will help physicians determine whether a patient is suffering from a migraine that can be treated using specific medications or whether the headache may be caused by a more serious issue.

She recently gave a poster presentation about her algorithm at the American Neurological Association meeting in Chicago on October 8. The algorithm contains the ID Migraine scale,⁵ which includes 3 questions: (1) Is there headache-related disability (eg, missing work or school)? (2) Are the headaches associated with light sensitivity? and (3) Is there nausea with the headaches?

The physician begins by asking the patient whether he or she experiences migraine headache. If the answer is yes, the patient is asked whether he has used a triptan in the past 24 hours. "Depending on whether or not they have already treated themselves at home, the decision tree goes either 1 of 2 ways," Friedman said. "Your hands are kind of tied if the patient already used a triptan, so you would go with an anti-emetic and an analgesic. If there is no relief, the physician should call neurology to admit the patient rather than keeping him in the ED for a long time or sending him home without relief."

If the patient is not a known migraine sufferer, then treatment decisions are based on age and certain red flags in the patient's history. "If the patient is older than age 50 years and is coming in with their first headache, we worry about giant cell arteritis, brain tumor, stroke, or hemorrhage," Friedman said. "The patient may need lumbar puncture or scan depending on his presenting symptoms."

Friedman says she is in the beginning stages of validating her algorithm and is tracking patients who have been to the ED for headache at her hospital. She records the data from a patient's preliminary diagnosis to final diagnosis and up through treatment and discharge.

After this retrospective phase, Friedman hopes to introduce the algorithm in a prospective study and determine whether the algorithm leads to changes in patterns of treatment, time spent in the ED, amount of money spent in the ED, and whether there is an impact on unnecessary scans performed in the ED. She said it will probably be a year before the testing will be finished and results will be presented.

WHAT ABOUT IHS CRITERIA?

The International Headache Society (IHS) diagnostic criteria is considered to be the most effective tool for the clinical diagnosis of headache.⁷ However, Friedman said the criteria are primarily used by headache specialists and for research. "It isn't really used in practice, and I would say most ED physicians have never heard of the IHS diagnostic criteria," she said.

That's why Vincent T. Martin, MD, professor of clinical medicine at the University of Cincinnati College of Medicine, recently developed a model for the diagnosis of migraine based on an abbreviated version of the IHS diagnostic criteria. As a primary care physician with an interest in the study of headache, Martin said he became interested in abbreviated migraine diagnostic criteria because he recognized that the diagnosis of migraine was often missed in primary care, and one of the main obstacles was the complexity of current IHS criteria. "I thought that if I simplified the diagnosis, then it would improve the recognition of migraine--particularly the atypical presentations of migraine such as bilateral headaches, nonthrobbing character, etc," he said.

To create his model, Martin and colleagues studied more than 1500 patients at 4 different practice sites. The researchers found that migraine can be effectively diagnosed using 2 separate, 3-variable models. Patients presenting with nausea and photophobia as well as pulsating pain or pain that grows worse with exertion were predictive of migraine headache (Figure 2).

COMPARING ALGORITHMS

Whether the screening tools for migraine assessment will impact the effectiveness of these algorithms is yet to be determined, but the differences may not be great, said Martin. "It is not surprising that a variety of combinations of questions are predictive of migraine," he said. "There are many ways to skin a cat." Nausea is the only common variable in the ID Migraine, POUNDing scale, and abbreviated IHS criteria for migraine diagnosis, which Martin says is the most predictive symptom.

"While we did find 3-variable models that were quite predictive of migraine, the single-variable model of 'nausea' was almost equally predictive. In fact, the presence of nausea conferred a probability of migraine of more than 80%, which is an excellent positive predictive value. I would recommend to physicians that if a patient complains of headache associated with nausea, and secondary headaches are absent, then he probably has migraine."

Friedman chose to incorporate the 3-variable ID Migraine scale into her algorithm because it is widely recognized and used at many institutions. According to validation studies, 93% of migraineurs have 2 of the 3 features, and the scale has 85% sensitivity and 86% specificity, she said.

However, in an interview with Applied Neurology, Booth said he felt the POUNDing criteria for the diagnosis of migraine were preferred to the ID Migraine scale. "While the POUNDing scale involved 5 distinct criteria, the ID Migraine scale included 3 variables, 1 of which was also part of the study entry criteria, thus making it a 2-variable screening tool." For these reasons, Booth and his colleagues feel the POUNDing criteria are more useful.

"We can all speculate about which screening tool is best--ID Migraine versus others--but until validated in the ED setting, it is a judgment call," Friedman said. "The ID Migraine scale was validated for use in primary care. If the Booth protocol is for use in the outpatient office versus the ED, the questions and time that the provider has to ask them may differ."

Martin is confident that the abbreviated IHS model will be most effective for the diagnosis of migraine. "It is one of the only models to use a verbal diagnostic interview, which more closely mimics an interview conducted by a physician in the office; for example, ID Migraine was developed as a written questionnaire," he said. "Also, the [abbreviated IHS] model performed well in those with medication overuse and chronic daily headache. I am unaware that other models were tested in those subpopulations, which represent some of the most difficult headache groups."

"Our study found that nausea and photophobia were the most predictive, and the ID Migraine study found the same thing," he said. "Therefore, I suspect that models that include those variables will perform better."

REFERENCES1. Blumenthal HJ, Weisz MA, Kelly KM, et al. Treatment of primary headache in the emergency department. Headache. 2003;43:1026-1031.
2. Michel P, Henry P, Letenneur L, et al. Diagnostic screen for assessment of the IHS criteria for migraine by general practitioners. Cephalalgia. 1993;13:54-59.
3. Lipton RB, Scher AI, Kolodner K, et al. Migraine in the United States: epidemiology and patterns of health care use. Neurology. 2002;58:885-894.
4. Lipton RB, Diamond S, Reed M, et al. Migraine diagnosis and treatment: results from the American Migraine Study II. Headache. 2001;41:638-645.
5. Lipton RB, Dodick D, Sadovsky R, et al. A self-administered screener for migraine in primary care: the ID Migraine validation study. Neurology. 2003; 61:375-382.
6. Detsky ME, McDonald DR, Baerlocher MO, et al. Does this patient with headache have a migraine or need neuroimaging? JAMA. 2006;296:1274-1283.
7. Headache Classification Subcommittee of the International Headache Society. The International Classification of Headache Disorders: 2nd edition. Cephalalgia. 2004;24:9-160.
8. Detsky ME, McDonald DR, Baerlocher MO, et al. Does this patient with headache have a migraine or need neuroimaging? JAMA. 2006;296:1274-1283.
9. Martin VT, Penzien DB, Houle TT, et al. The predictive value of abbreviated migraine diagnostic criteria. Headache. 2005;45:1102-1112.