Burden of Migraine in the United States:  Disability and Economic Costs FREE

APPROXIMATELY 6% of men and 18% of women in the United States currently suffer from migraine headaches.¹ Despite this high prevalence and the disabling effects of migraine attacks, migraine has not received adequate attention as a public health priority because its impact on society has been underestimated. In the American Migraine Study, a total of 71% of men and 59% of women determined to have migraine from self-reported symptom data had never received a diagnosis from a physician.² While most migraine sufferers reported taking medication for headache, most used over-the-counter preparations. Even among migraine suffers with moderate or severe headache-related disability, less than half were treated with prescription medication.²

In recent years, a number of reports^3- 13 have described the burden of migraine disease.Two of the studies^3,5 have estimated the burden of the disease in the United States but results may be unreliable due to the limitation of study design. One study³ underestimated the prevalence of migraine relative to population-based studies by using self-reported migraine rather than systematic ascertainment of migraine. Another study⁵ overestimated attack frequency and disability relative to the population by using biased samples composed of participants in controlled clinical trials. In this study, we estimate direct and indirect migraine-related costs using population-based estimates of migraine prevalence, attack frequency, disability, and health care resource use.

Although the overall cost attributable to migraine is difficult to measure, most of the impact has been associated with disability and decreased functional status and consequent indirect costs to employers.^5,7- 8 Consistent with this view, we have developed a model for the US population from a societal perspective based on information from the literature supplemented with additional analyses of medical claims.

To estimate migraine burden of illness, 4 estimates were derived: bedridden days per year, health care resource use, economic loss due to missed workdays, and economic loss due to reduced productivity. Health care cost estimates include the population 5 years of age or older; other estimates only apply to the population from 20 to 64 years of age. Data input and their sources are summarized in Table 1.

We estimated the number of migraine sufferers based on the projected US population distribution in 1998 by the Census Bureau¹⁴ and the age- and sex-specific prevalence rates of migraine as reported from studies that used a definition of migraine consistent with the criteria set forth by the International Headache Society (IHS).¹⁹ The IHS criteria were based on the consensus of experts; they are explicit and have been widely used in both population-based surveys and clinical trials. Prevalence rates for adolescents and adults aged 15 years or older were obtained from the American Migraine Study,¹ which is a nationwide population-based survey covering 15,000 households. The prevalence rates in children and preadolescents (aged 5-14 years) were estimated from a French study,¹⁶ which assessed the prevalence of migraine in a random sample of the entire childhood population of a single city, also using the IHS criteria.

Migraine-related disability was calculated as a function of the number of bedridden days in patients aged 20 to 64 years based on results of the Baltimore County Migraine Study.¹⁵ Subjects reported how often they needed bed rest when experiencing a migraine attack; response options were: never, rarely, less than 50% of the time, and more than 50% of the time; for calculations, these were translated into rates of 0%, 10%, 25%, and 75%. An average percentage of attacks in which patients need bed rest (PAB) was generated for each age and sex stratum. Total number of bedridden days per year (BDY) in the given age and sex stratum was calculated as follows:

,

where NMS is the number of migraine sufferers in an age and sex stratum; FAY, the frequency of migraine attacks per year; and ABH, the average bedridden hours when lying down with a migraine attack. By dividing by 24, we convert bedridden hours into bedridden days. The Baltimore County Migraine Study, from which bedridden hours were derived, explicitly excluded regular sleeping hours. These data elements are summarized in Table 2.¹⁵

The costs of health care resource utilization associated with migraine were obtained from an analysis of 1994 data from MEDSTAT's MarketScan data set. This includes both inpatient, outpatient, and prescription drug claims for employees and their dependents for more than 40 large employers in the United States. For our analysis, we included only enrollees with continuous coverage for 1994, which was about 1 million lives. The prevalence of migraine (diagnosed and undiagnosed) for each age and sex stratum in this population was assumed to be the same as in the general US population. Charges in the MEDSTAT database were used as cost inputs to the model. The model was restricted to migraine-associated health care costs.

Migraine-related medical encounters were identified through matching International Classification of Diseases, Ninth Revision, Clinical Modification²⁰ codes in diagnostic fields. Migraine-related medical visits were defined by encounters with the first 3 digits 346 (migraine), or the combination of prescriptions for migraine-specific drugs associated with an encounter coded with 784.0 (headache) or 307.81 (tension headache). For outpatient encounters, either primary and secondary diagnoses were accepted. For hospitalizations, only the principal diagnosis was accepted. The migraine-specific medications are ergotamine tartrate (Ergomar, Ergostat), ergotamine tartrate combinations (Cafetrate, Ercaf, Migergot/Migergot-PB, Wigraine, Bel-Phen-Ergot, Bellergal/Bellergal-S, Phenerbel-S, Cafergot/Cafergot-PB, Ergo-Caff/Ergo-Caff-PB), dihydroergotoxine (Ergoloid Mesylates), dihydroergatamine (D.H.E. 45), sumatriptan (Imitrex), and methysergide (Sansert).

Migraine-related drug costs were estimated only for those patients who had at least 1 migraine-related medical encounter as defined earlier. All abortive medications classified as migraine medication were included, ie, those with Generic Product Identifier rubric starting with 67. Generic Product Identifier categorizes drug products by a hierarchical therapeutic classification scheme developed by Medi-Span Inc,²² which was linked to the National Drug Code¹⁸ as recorded in the MarketScan database. We did not include prophylactic drugs other than methysergide (Sansert) in the estimate of migraine-related drug costs because such drugs are commonly used to treat conditions other than migraine.

The impact of migraine on employers was evaluated as missed workdays and impaired work performance¹; these measures were combined with data from the Bureau of Labor Statistics with respect to percentage of population working for pay and average working hours per week.¹⁷ The total number of migraine-related missed workdays (TMWD) per year was calculated for each age- and sex-specific stratum as follows:

,

where MWD is average migraine-related missed workdays per year; based on the previous research, we applied 3.8 and 8.3 days for male and female employees in the model²³; PWP, percentage of the population working for pay, estimated at 73% for males and 57% for females; and WHW, average working hours per week, most recently reported as 35 working hours for both sexes.¹⁷ By dividing by 40, we convert the working hours into the standard full-time level (8 working hours per day).

Impaired work performance was calculated as a function of the number of workdays with migraine (NWDM) and reduced work efficiency during the attacks. The NWDM was estimated as follows:

The average number of workdays with migraine (WDM) per year was estimated based on patient self-report.²² For patients who reported staying at work with migraine at a frequency of 1 to 2, 3 to 5, 6 to 8, and 11 to 20 days over a 1-year period, we used the median value of each category, and for patients who reported staying at work for 21 or more days, we assumed that it was 25 days. The average workdays with migraine per year were estimated at 7.5 days for men and 7.6 days for women.

Lost workday equivalent (LWDE) due to impaired work performance was calculated as follows:

,

where EWM is the average effectiveness at work with migraine, estimated at 42% for men and 34% for women.²²

The total employment lost due to migraine (TELM) in dollar terms was calculated as follows, assuming 8 hours for each working day:

.

The national nonfarmer average hourly salary ($12.09 in February 1997) was used in the model.¹⁸

Table 3 displays the projected number of migraine sufferers aged 20 to 64 years in the United States in 1998 based on US Census projections and prevalence data from the American Migraine Study. Migraine-related bedridden days are presented in Table 2. The average number of attacks per year was 34 for men and 37.4 for women. Overall, about 58% of migraine attacks needed some bed rest. While younger migraineurs were more likely to be restricted to bed during the attacks, older women tended to stay in bed for longer duration. Female patients spent 6 hours in bed on average during attacks requiring bed rest compared with 4.5 hours in male patients. On an annual basis, migraineurs were restricted to bed 3.8 days for men and 5.6 days for women. For the US population, this results in 112 million migraine-related bedridden days per year.

The direct medical costs associated with migraine treatment are displayed in Table 4. We identified over 1 billion health care dollars in annual treatment costs in the United States. Female patients accounted for about 80% of the costs. Expenses related to physician office visits accounted for about 60% of all costs, while prescription drug accounted for near 30%. Notably, emergency department costs attributable to migraine were less than 1% of the total costs in both sexes. Assuming 29% male and 41% female patients were medically diagnosed,² we estimated that close to $100 were spent per diagnosed patient per year.

Table 5 presents estimated labor costs broken down by migraine-associated days off work and reduced work performance. The indirect costs to American employers were estimated at approximately $13 billion annually. Assuming that the percentage of people working for pay is the same for migraineurs as for the general population (ie, 73% of males and 57% of females), this breaks down to $690 and $1127, respectively, for each male and female in lost productivity costs. Most indirect costs (almost $8 billion) are due to migraine-related missed workdays. Female patients consistently incurred higher costs in both workday loss and reduced function at work. Similar to the migraine-related treatment costs, female patients accounted for about 80% of total labor costs due to migraine. Patients aged 30 to 49 years in both sexes incurred higher indirect costs compared with younger or older employed patients.

Patients with migraine are frequently disabled during their acute attacks. Although migraine-related disability can be reflected by both bedridden days and restricted activities, we emphasized bedridden days in our calculations because they can be more reliably reported and quantified. Results of both the American Migraine Study and a population-based survey in Canada indicate that about one third of migraine sufferers experienced severe disability or the need for bed rest following attacks, and an additional 50% reported mild or moderate disability.^1,23 While the percentage of patients with migraine who need bed rest at attacks is striking, this measure of the burden of disease has not been well quantified previously.

Using data from the National Health Interview Survey, Stang and Osterhaus³ reported that patients with migraine spent more than 3 million bedridden days per month, which is about one third of our estimate of 112 million migraine-related bedridden days per year in the United States. The discrepancy in the estimates is almost entirely because the migraine prevalence rate used in their study was only 4.1%. In the National Health Interview Survey study, diagnosis of migraine was based on patients' self-report. As a result, only those patients who had been previously diagnosed by physicians related their bed rest to migraine attacks. Migraine prevalence rates used in our analysis were estimated using operational diagnostic criteria proposed by the IHS, and thereby included the entire migraine population. Estimates from the American Migraine Study²⁴ are in line with other major population-based studies. Available data have consistently shown that no more that half the migraine sufferers who met IHS's criteria had received a diagnosis by physicians,^2,25 accounting for the discrepancy between prevalence based on the American Migraine Study and the National Health Interview Survey.

Our analysis suggests that the mean length of bed rest due to migraine attacks was 4.5 hours for men and 6.0 hours for women. Of interest was the finding that severity of migraine attacks, as measured by the proportion of attacks that require bed rest, tends to decrease with increasing age. The length of bed rest in female patients aged 50 to 64 years was nearly twice that of those aged 20 to 29 years. Another intriguing result was that the same proportion of men and women required bed rest; however, women required longer stays in bed in all age groups. Other studies have reported sex differences in the severity of migraine. Women generally report more intense pain and more headache-related disability than men.²⁶

Direct costs due to migraine-related medical care in the United States have not been extensively studied. Most studies^3,27 have only reported health care resource use, without assigning monetary costs, making it difficult to compare the economic burden of migraine with that of other chronic disorders. Using a managed care claims database, Clouse and Osterhaus⁴ revealed that migraineurs generated nearly twice as many medical claims as comparable group patients, and nearly 2.5 times as many pharmacy claims. The average payment per member-month of enrollment was $145 for migraineurs vs $89 for other patients. However, the analysis did not differentiate migraine-specific costs from the total medical expenses. Since migraines are known to be associated with several neurologic and psychiatric conditions, including depression, epilepsy, and stroke,^28- 31 the excess costs of migraineurs were partly attributable to other coexisting disorders. Osterhaus et al⁵ surveyed 648 patients and concluded that direct migraine-related medical costs were $817 per year, which is significantly higher than the estimate from our model. We believe that the direct health care costs were overestimated in their study due to its selection of a study population drawn from enrollees in migraine clinical trials. Nevertheless, this suggests that patients with the most severe migraine may account for a disproportionate share of total health care expenses. In a managed care–based study, the cost of headache care was shown to increase with the severity of clinical disease, supporting this hypothesis.⁸

Migraine-related indirect costs are often quoted from 2 US studies.^3,5 Using data from the National Health Interview Survey, Stang and Osterhaus³ estimated that migraine-associated lost productivity was valued at $1.4 billion per year, which is far less than $13.3 billion in our estimate. As noted earlier, only medically diagnosed patients were included, which was estimated at 6.2 million patients. In our study, we estimated that there would have been 21.8 million migraine patients aged 20 to 64 years in 1998, and this was the basis for our estimates of indirect costs.

Osterhaus et al⁵ estimated that annual lost labor costs due to migraine ranged from $5.6 to $17.2 billion. It is worth noting that although the total indirect costs from our estimate are in the same range, the costs per employed migraine patient per year were much higher in their study at $6864 for men and $3600 for women. In our view, the underestimate of prevalence balanced the overestimate of cost per migraineur. Their study population reported an average of 26 missed workdays and 70 days working with migraine per year. These figures drawn from clinical trial participants are significantly higher than estimates from population samples.^6,22 We estimate annual means for men and women, respectively, of 3.8 and 8.3 missed workdays and 7.5 and 7.6 migraine-impaired working days. On the other hand, the earlier study used much lower prevalence rates (1.8% for men and 5.2% for women); our higher prevalence estimates were comparable with most population-based studies in the United States and Canada.^25,32

Our estimates of lost labor costs due to migraine are based on projections from epidemiological survey data. The projections do not capture several important components of the burden of disease; we do not measure unemployment and underemployment due to migraine.⁸ Nor do we capture the value of lost homeworker time for chores, family, social, and leisure activities. We do not measure other aspects of disease burden. For example, a parent may miss work to care for a child with migraine or to make a medical visit for themselves or a family member. Finally, we do not capture the substantial burden that may exist between attacks.⁸ For all these reasons, we considerably underestimate the true burden of disease.

Our analysis suggests that direct costs represent a small proportion of the overall societal costs of migraine. Nevertheless, there are a few reasons to believe that the direct costs have been underestimated. First, our analysis of medical claims does not capture all migraine-related treatment costs because the disease is often not treated with specific therapies. Second, we did not measure over-the-counter or preventive medications and nondrug-related interventions. Third, although it is desirable to adjust direct costs to reflect 1998 dollars, we opted not to make an arbitrary adjustment such as using the consumer price index because of uncertainty about changes in migraine-related costs. Migraine-related health care resource utilization may have increased significantly since 1994, the year we analyzed in the MEDSTAT database. The IMS reports that annual antimigraine medication sales increased from $86.2 million for the 12 months ending April 1993 to $707.1 million for the 12 months ending April 1997.³³ The IMS data for 12 months ending April 1994 was about $250 million, which is reasonably close to our estimate of annual drug costs ending in December 1994 of $300 million. Finally, MEDSTAT represents an insured population sponsored by major employers, which may not be representative of the population in other settings.

In conclusion, the results of this study show that the burden of migraine disproportionally falls on patients and their employers. Third-party payers shoulder less than 10% of migraine-related economic costs. The overall cost of migraine to society is large, and comparable with that reported for diabetes and higher than that reported for asthma.^34- 35 Various effective treatment regimens are now available and their impact on the reduction of the disease burden needs to be further evaluated in different settings.

Accepted for publication July 21, 1998.

Reprints: X. Henry Hu, MD, MPH, PhD, PO Box 4, WP39-164, West Point, PA 19422.