Abstract
Background: The Migraine Disability Assessment (MIDAS) instrument is a five-item questionnaire developed to measure headache-related disability and improve doctor-patient communication about the functional consequences of migraine.
Objectives: To examine the test-retest reliability and internal consistency of the five items and of the overall MIDAS score in population-based samples of migraine sufferers in two countries and to compare consistency across countries.
Methods: Using a clinically validated telephone interview, population-based samples of migraine-headache sufferers were identified in the United States (Baltimore, MD) and the United Kingdom (Merton and Sutton, Surrey). Eligible individuals completed the MIDAS questionnaire on two occasions an average of 3 weeks apart. The MIDAS score is derived from five questions about missed time from work and household work (one question each about missed days and days with at least 50% reduced productivity) and missed days of nonwork activities.
Results: A total of 97 migraine-headache sufferers from the United States and 100 from the United Kingdom completed the MIDAS questionnaire twice. Mean and median item values and overall MIDAS scores were similar between the United States and the United Kingdom. Test-retest Spearman correlations of individual items ranged from 0.46 to 0.78. No significant differences in item-specific correlations were observed between the United States and United Kingdom. The test-retest Pearson correlation of the MIDAS score (i.e., sum of lost days and reduced-effectiveness days in each domain) was 0.80 in the United States and 0.83 in the United Kingdom. The Cronbach α, a measure of internal consistency, was 0.76 in the United States and 0.73 in the United Kingdom.
Conclusions: This is the first international population-based study to assess the reliability of a disability-related illness severity score for migraine. The reliability and internal consistency of the Migraine Disability Assessment score are similar to that of a previous questionnaire (Headache Impact Questionnaire). However, the Migraine Disability Assessment score requires fewer questions, is easier to score, and provides intuitively meaningful information on lost days of activity in three domains.
Treatment choices for headache should reflect both the headache diagnosis and the overall severity of the disorder.1,2 Although standardized criteria are available for headache diagnosis, assessing the severity of a headache condition is not standardized and poses a number of challenges. Headaches vary in their individual impact from mild pain with no disability to severe pain with prolonged incapacitation.1,2 Within individuals, headaches vary from attack to attack and over time in frequency of occurrence and in intensity. Finally, the severity of a headache condition must be based on the disability caused by attacks occurring over a defined, clinically relevant period and not on the disability from any one attack. To address these challenges in measuring severity, we focused on assessing disability that is a consequence of headaches occurring over a defined interval.
In measuring the functional consequences of illness, the World Health Organization distinguishes impairment, functional limitations, and disability.3 Impairment refers to the primary manifestations of illness, such as pain or a limited range of motion. Functional limitations are defined by the activities a person cannot do, such as walking. Disability refers to the consequences of illness on work and function in other roles. We have selected disability as a focus for measuring headache severity because it reflects the true functional consequences of an illness such as migraine.
A previous study, the Headache Impact Questionnaire (HImQ), was developed to measure headache severity.4,5 The HImQ score was derived as the sum of pain intensity and total lost time in each of three activity domains (work for pay, housework, and nonwork activities). Although the HImQ was highly reliable4 and was suitable for research, the calculation formula for the HImQ score was too complicated for self-scoring or use in primary care. The score was based on evidence that pain and disability are hierarchically related and that they can be scaled together.6,7 Although this concept has abundant psychometric support, pain and lost time are scaled in different units. Consequently, a composite score does not have intuitively meaningful units, potentially limiting its acceptability in primary care. We found that eliminating the pain component of the HImQ score did not substantially alter either the reliability coefficient or the relative score.
To simplify scoring and provide a score with interpretable units, we developed the Migraine Disability Assessment (MIDAS) questionnaire, a simple 5-item self-administered questionnaire designed to capture information on lost time from work for pay, housework or chores, and nonwork activities. The reason for developing MIDAS was that if it proved to be reliable and valid, it might serve as a screening tool to identify individuals in need of medical care and as an outcome measure in clinical practice, clinical trials, and epidemiologic research. The purpose of this study was to examine the test-retest reliability of the MIDAS score in population samples of migraine sufferers in the United States and United Kingdom.
Methods.
In this section, we describe the rationale for the MIDAS questionnaire and its relation to previous work3,4 followed by the methods for the population-based surveys in the United States and United Kingdom and the reliability studies. There were two distinct phases to the study in each country. In the first phase, we conducted a population-based telephone survey within defined catchment areas in the United States and United Kingdom. The methods for selecting individuals for interview from the population were similar to those used in a previous study8 and are summarized below. In the second phase, migraine-headache sufferers identified from the telephone interview were enrolled in a study to assess the test-retest reliability of the MIDAS questionnaire.
MIDAS questionnaire development.
The rationale for the development of the MIDAS questionnaire (see Appendix) has been described in detail elsewhere.9 Like the HImQ, the MIDAS questionnaire captures information on disability in terms of missed days at work for pay, household chores, and nonwork time. To capture information on reduced productivity, the following question was asked: “On how many days in the last 3 months was your productivity at work or school reduced by half?” This question format captures the number of days with substantial productivity reduction. Although these are counted as full lost days, days during which productivity is moderately reduced (i.e., <50%) are not counted. In a separate diary validity study, measurement penalty or bias for this simplification is being examined. This question format was also used for household chores. Because individuals had difficulty interpreting this type of question for nonwork activities in pilot studies, we excluded it.
The MIDAS score was derived as the sum of lost days due to headache as follows: one question about the extent to which headaches interfere with nonwork activity (MISS-LEISURE) and two questions each about work (MISS-WORK + WORK HALF) and work at home (MISS-CHORE + CHORE HALF).
Population sample and survey.
The population surveys were conducted in Baltimore, Maryland, and in the boroughs of Croydon, Kingston, Sutton, and Merton within greater London in the United Kingdom. These areas were selected because they are demographically diverse with regard to age, race, and socioeconomic status.
The method for selecting individuals is similar to that used in a previous survey.8 In brief, a representative sample of households was selected from both populations using random digit telephone dialing methods. Ten attempts were made to contact each selected household. At the time of initial telephone contact, a census of the household was obtained to identify all age eligible individuals (i.e., between the ages of 18 and 55 years). To obtain a census, the gender and age of each household member was ascertained. Before proceeding to the interview, oral informed consent was obtained, and the purpose of the survey was described to the respondent. Individuals were told that the interview would require 5 to 15 minutes to complete. Respondents who agreed to participate were interviewed about their headaches using a clinically validated computer-assisted telephone interview (CATI).3 A total of 5,769 interviews were completed in the UK boroughs for a participation rate of 70.9%; a total of 7,275 interviews were completed in Baltimore, Maryland, for a participation rate of 67.8%.
The CATI has been described in detail elsewhere.8 Participants were first asked if they had at least one headache not due to a head injury, hangover, pregnancy, or an illness such as a cold or flu. Detailed questions were asked about up to two different self-defined types of headache that occurred in the last 12 months, including all diagnostic features of migraine specified by the International Headache Society (IHS).10 Each headache feature reported in the affirmative (ever versus never) was followed by a question about how often the feature occurred with the specific headache type. The response options were as follows: never, rarely, less than half the time, and half the time or more. When nausea, photophobia, or phonophobia was reported in the affirmative, a follow-up question was also asked about the severity of the symptom (mild, moderate, or severe).
The validity of a CATI based diagnosis of migraine was assessed in the United Kingdom. A random sample of individuals who met CATI criteria for migraine (n = 61) and a sample who did not (n = 50) were independently evaluated by two clinicians who used a semi-structured interview for headache diagnosis,11 and followed IHS criteria in assigning a diagnosis.10 The diagnostic sensitivity of the CATI was 91%, and the specificity was 97%. The positive predictive value (i.e., proportion of individuals identified from the CATI as having migraine who actually have migraine) was 94%. The screening validity of the CATI for migraine in the UK study was similar to results observed in an earlier study in the United States.8
Reliability study.
Individuals who completed the telephone interview and who received a CATI diagnosis of migraine headache were invited at the time of the interview to participate in the reliability study. Those who gave oral consent received the MIDAS questionnaire by mail with a letter of explanation. Participants were asked to return the questionnaire in a postage-paid self-addressed envelope. After a subject returned the first copy of the MIDAS questionnaire, a second copy was sent approximately 18 days later.
We selected sufficient migraine-headache sufferers so that approximately 100 individuals would complete the MIDAS questionnaire twice, with approximately 3 weeks between mailings of the questionnaires. In the United States, 179 migraine-headache sufferers were invited to participate in the reliability study. A total of 124 agreed to participate. Of the 124 who were sent an initial MIDAS questionnaire, 109 (87.9%) returned the questionnaire. Of the 109 sent a second questionnaire, 97 (89.0%) returned the questionnaire. In the United Kingdom, a total of 154 migraine-headache sufferers were invited to participate in the reliability study. A total of 141 agreed to participate. Of the 141 who were sent an initial MIDAS questionnaire, 117 (83.0%) returned the questionnaire. Of the 117 sent a second questionnaire, 100 (85.5%) returned the questionnaire. Individuals were paid a total of $5 (£5 in the United Kingdom) for taking the time to complete and return each MIDAS questionnaire.
Analysis.
The goal of the analysis was to evaluate the reliability of responses to specific questions and, more importantly, the reliability of the overall MIDAS score. Analysis included frequency distributions and plots to identify outliers and Spearman and Pearson correlations to assess reliability between responses to the first and second questionnaires. The Spearman correlation tends to be conservative because it is not usually influenced by outliers, whereas the Pearson correlation coefficient is influenced by outliers. Internal consistency of the MIDAS scoring was assessed using the Cronbach α, a measure that is analogous to a split half reliability assessment. The Cronbach α can range from 0 to 1.0. An α of 0.7 is considered acceptable; an α of 0.8 or greater indicates that the internal consistency of the scale is excellent.
Results.
We first compared individuals with migraine in the United States and in the United Kingdom, who completed both MIDAS questionnaires, with the respective population samples of migraine cases. Results of the reliability analysis follow.
Study participants versus all migraine sufferers in the population.
The 97 individuals in the United States and the 100 individuals in the United Kingdom who completed both MIDAS questionnaires differed on demographic characteristics, but not headache characteristics, when compared with the total population of migraine cases identified in the telephone surveys (table 1). Compared with all migraine cases, reliability participants in the United States and United Kingdom were slightly more likely to be female and older.
Percent distribution of individuals with migraine who completed both Migraine Disability Assessment (MIDAS) questionnaires by demographic and headache features obtained during the baseline telephone interview compared with the sample of all population-based individuals with migraine eligible for the US and UK reliability studies, 1997–1998
Compared with all migraine sufferers in the United States, reliability participants had more frequent headaches, headaches of longer duration, and more symptoms. In the United Kingdom, reliability participants had less frequent headache than all population-based cases but were similar to all migraine cases on other headache features. Overall, reliability study participants in both countries had headaches of longer duration.
Evaluation of the MIDAS items.
Using data from the first completed MIDAS questionnaire, migraine sufferers had frequent headaches in a 3-month period (in the United States: mean = 14.9 and median = 10; in the United Kingdom: mean = 10.7 and median = 7) with relatively high levels of pain as rated on a 0- to 10-point scale. On average, in the population sample headache frequency was higher among US migraine cases compared with those in the United Kingdom. The higher frequency of headaches among US reliability study participants reflects both this difference in the epidemiology of attack frequency between the United States and the United Kingdom and selective participation of individuals with more frequent headaches in the United States compared with the United Kingdom.
The number of days that migraine sufferers reported disability differed by activity domain. In both the United States and the United Kingdom, more missed days were reported for household work followed in order by nonwork activities and school or work for pay (table 2). Days during which productivity was reduced by half or more at work or school were considerably more common than missed work or school days. In contrast, missed household work days were similar to days during which productivity was reduced by half in both the United States and the United Kingdom.
Mean and median responses to Migraine Disability Assessment (MIDAS) questions from participants in the UK and US test-retest reliability studies overall and among participants reporting at least six migraine headaches per year, 1997–1998
The overall mean MIDAS score was significantly higher in the United States than it was in the United Kingdom (see table 2). Median scores were also higher in the United States. This difference, however, is confounded by differences in headache frequency between US and UK study participants. We explored these differences in linear regression analysis, where MIDAS score was the dependent variable and country (United States versus United Kingdom) and headache frequency (linear and quadratic terms) were independent variables. The headache frequency adjusted difference in mean MIDAS score between the United States and the United Kingdom was 2.91 (95% confidence intervals of −0.40 to 6.22), a difference that was not statistically significant (p = 0.09).
Test-retest reliability.
Overall, migraine sufferers were reliable in reporting information on lost time associated with headaches (table 3). The test-retest Pearson correlation ranged from 0.54 (reduced productivity at work) to 0.68 (missed work days) in the United States and from 0.52 (missed nonwork days) to 0.82 (missed days of household work) in the United Kingdom. Values for Spearman correlations were similar to Pearson correlations, suggesting no substantial influence by outliers.
Pearson and Spearman correlation coefficients for responses to Migraine Disability Assessment (MIDAS) questions and the overall score among participants in the US (n = 97) and UK (n = 100) test-retest reliability studies, 1997–1998
For the overall MIDAS score, the test-retest correlations were relatively high, clustering around 0.80, and similar between the United States and the United Kingdom. The Cronbach α, a measure of internal consistency, was 0.76 in the United States and 0.73 in the United Kingdom.
Discussion.
The MIDAS questionnaire was developed to provide a brief disability tool with intuitively meaningful units and simple scoring rules to make it practical for use in primary care. In developing the MIDAS questionnaire, we hoped to facilitate communication about headache-related disability in clinical settings, to provide a self-assessment procedure for screening individuals with severe disease who could benefit from medical care, and to provide an outcome measure for clinical practice, clinical trials, and epidemiologic studies. In previous research, we found that a related instrument, the HImQ, was highly reliable (Pearson correlation of 0.85 and Cronbach α of 0.83)4 and valid.5 However, the overall impact score was difficult to calculate and interpret.
The MIDAS score is based on responses to five questions about disability associated with headache in the previous 3 months. Responses to each question are scaled in units of number of days missed (work or school, household work, and nonwork activities) or the number of days during which productivity was reduced by half or more (work or school and household work) in the past 3 months. The test-retest Pearson correlation for MIDAS was similar to the HImQ, even though fewer items were used to derive the MIDAS score. Moreover, the scoring algorithm for MIDAS was determined a priori, whereas the scoring algorithm for the HImQ was developed after completing factor analysis and only used 8 of the original 16 items. The exploratory nature of the analysis in developing a scoring algorithm for the HImQ may have lead to slightly optimistic reliability.
The reliability of the MIDAS score is likely underestimated because there was an average of 21 to 22 days between completed questionnaires. This time interval represents 25% of the 3-month recall period. Random variation in different time intervals as well as real changes in some individuals over this period could have contributed to differences in scores derived from the first and second MIDAS questionnaires.
For MIDAS, we worded questions differently to both simplify scoring and to approximate a measure of lost day equivalents without the need for multiplication. Although this simplification may reduce accuracy, we believe this most likely introduces a small amount of error. We overestimate lost time by counting days with reduced productivity of 50% or greater as if the entire day were lost. But we do not count days during which productivity is reduced by less than 50%, leading to underestimation of lost time. Because these approximations lead to overestimation and underestimation, they partially offset each other. The MIDAS score may under- or overestimate true disability in patients who experience relatively fixed levels of disability from attack to attack. By counting days, we may overestimate disability time if an individual misses activity for part of a day but counts it as a whole day. We are currently evaluating whether meaningful bias does occur in a separate validity study using a daily diary to measure disability.
Because the MIDAS score separately assesses lost time in each of three domains (work, chores, nonwork activity), a single day of disabling headache may contribute more than one point to the total score, depending upon the duration and severity of the attack and range of scheduled activities for that day. Our measurement strategy assumes that a day is typically divided into segments corresponding to different domains of activity. If an individual loses a day of work and recovers in time to engage in other activities after work, this adds only one point to the MIDAS score. If both work and nonwork activities are missed, this adds two points to the MIDAS score. The MIDAS score therefore reflects the range of intended activities lost on a particular headache day summed over a 3-month period. In part, the overall MIDAS score also reflects an established pattern of the relative influence of headache on each domain of activity. In this study and previous studies4,5 we showed that headache sufferers are more likely to miss chores than nonwork activity (family, social, leisure) and are more likely miss nonwork activities than work. A very severe headache is therefore likely to result in missed activity in more than one domain. Because a headache day can contribute up to three points to the MIDAS score, scores must be interpreted in light of the total number of days with headache pain. This information is provided in question A on the MIDAS questionnaire.
The MIDAS severity measure is similar in some respects to quality of life (QOL) measures but differs in important ways. Like QOL measures, the MIDAS is intended to capture the aggregate impact of illness on an individual over a period of time. Because MIDAS focuses on domains of life thought to be most strongly influenced by migraine, it has more in common with disease-specific12-16 rather than generic17-20 instruments for assessing QOL.
The MIDAS questionnaire differs from QOL instruments in design and content. QOL instruments measure aspects of an individual’s health status over a period of time. For episodic disorders such as migraine, this includes both days with and days without headache. MIDAS focuses on days with headache and is intended to measure severity of illness, expressed as a composite of lost time in various domains, not QOL. The MIDAS severity measures combine items that were known, a priori, to be hierarchically related. For example, pain often occurs in the absence of limitations to activities, and lost time in chores and nonwork activities occurs, on average, at a lower threshold than lost work time.4-6 In contrast, QOL instruments are used typically to derive domain-specific measures (of physical or social role function, for example) and a global measure that is a composite of all domain-specific measures. The manner in which items are hierarchically related is not explicitly considered in devising a QOL scaled score.
QOL instruments tend to be much longer than the MIDAS and have arbitrary or percentile scores. We hope that the brevity and the use of intuitively meaningful units will increase the acceptance of the MIDAS questionnaire in primary care.
This study has a number of strengths. First, because the studies were population based, the full spectrum of migraine headache sufferers was represented. Second, because parallel studies were conducted in the United States and the United Kingdom, the demonstration of reliability is international in scope. The MIDAS questionnaire is based on extensive prior work, is highly reliable, easy to use, and intuitive, and provides an explicit measure of lost time. Finally, the reliability has been demonstrated in two independent studies in two separate countries. In future work, the validity of the MIDAS score has to be demonstrated. It is important to note that the MIDAS questionnaire is not a diagnostic tool nor a substitute for a diagnostic tool. It is intended to complement diagnostic assessment and facilitate treatment decisions. Clinicians need to make a diagnosis and assess disability as a prelude to treatment.
The MIDAS questionnaire
Instructions: Please answer the following questions about ALL the headaches you have had over the last 3 months. Write your answer in the box next to each question. Write zero if you did not do the activity in the last 3 months. (Please refer to the calendar below, if necessary.)
1. On how many days in the last 3 months did you miss work or school because of your headaches?
2. How many days in the last 3 months was your productivity at work or school reduced by half or more because of your headaches (do not include days you counted in question 1 where you missed work or school)?
3. On how many days in the last 3 months did you not do household work because of your headaches?
4. How many days in the last 3 months was your productivity in household work reduced by half or more because of your headaches (do not include days you counted in question 3 where you did not do household work)?
5. On how many days in the last 3 months did you miss family, social, or leisure activities because of your headaches?
A. On how many days in the last 3 months did you have any headache (if a headache lasted more than one day, count each day)?
B. On a scale of 0 to 10, on average how painful were these headaches (0 = no pain at all, and 10 = pain is as bad as it can be)?
Innovative Medical Research, Inc.
Acknowledgments
Supported by Zeneca Pharmaceuticals.
Acknowledgment
The authors thank Michael Von Korff for collaboration on earlier work and the MIDAS Advisory Committee for their thoughtful contributions (Drs. John Edmeads, Michel Ferrari, Peter Goadsby, Helen Massiou, Ninan Mathew, Alan Rapoport, and Steve Silberstein).
- Received January 27, 1999.
- Accepted April 10, 1999.
References
Disputes & Debates: Rapid online correspondence
NOTE: All authors' disclosures must be entered and current in our database before comments can be posted. Enter and update disclosures at http://submit.neurology.org. Exception: replies to comments concerning an article you originally authored do not require updated disclosures.
- Stay timely. Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- 200 words maximum.
- 5 references maximum. Reference 1 must be the article on which you are commenting.
- 5 authors maximum. Exception: replies can include all original authors of the article.
- Submitted comments are subject to editing and editor review prior to posting.