This study aimed at validating mMIDAS and studying the QoL of a sample of migraine patients. Under no circumstances, it interfered with health professionals’ decisions concerning the most appropriate medical approach for each patient. Before data collection, all participants filled in a written informed consent and this study received the approval from the Ethics Commission of the Coimbra University Hospital.
The study was conducted between September 25th and December 6thth 2019 at the Neurology Department of Coimbra University Hospital. Studied population consisted of consecutive adult patients observed in a headache outpatient clinic. These were evaluated and included in the study by a clinician if older than 18 years, with chronic or episodic migraine, with ability to give consent to participate in the study, and if able to read and write Portuguese for self-completion of questionnaires. Excluded were unstable patients or with uncontrolled symptoms and considered by clinicians as unable to fill the measures, cognitively affected, and those who could not understand Portuguese. The questionnaire was completed during the consultation.
Besides mMIDAS, we applied a generic QoL instrument (EQ-5D-5L), two questions about satisfaction with life and social support, a specific psychological instrument to measure anxiety and depression (HADS); and a short socioeconomic set of questions.
- mMIDAS is a self-administered questionnaire that contains seven questions about the headache a patient had in the previous month. The first five questions assess the impact of migraine on three domains of daily activity: two questions for paid work or schoolwork, two questions for household work, and one question for family, social and leisure activities. The two questions for each of the first two groups assess, respectively, the number of days off due to headache, and the number of days in which the productivity was reduced by half or more.
The sixth and seventh additional questions, for the clinicians benefit, count the frequency of headache days in the previous month, and the average intensity of the headache attacks, in a 0 (no pain at all) to 10 (pain as bad as it can be) scale.
mMIDAS index is derived from the sum of the answers on the first five questions. This score defines patients in four categories of headache disability: little/none disability if the score is between 0 and 5; mild disability if between 6 and 10; moderate disability if between 11 and 20; and severe disability if greater than 20 [4-6].
A first Portuguese non-validated translation of mMIDAS has been provided by the author through Mapi Research Trust (see Supplementary file). Therefore, before implementing this measure, we decided to validate its contents by clinical reviews with Portuguese neurologists and by cognitive debriefings with patients. We also tested its reliability, as well as its construct and criterion validity.
- EQ-5D-5L is a short generic QoL preference measure that allows generating an index representing the value assigned to an individual's health status. It was developed by the EuroQoL Group in 1987 and it is currently composed of a descriptive system and an visual analogue scale (VAS). The descriptive system represents health in five dimensions: mobility, self-care, usual activities, pain/discomfort and anxiety/depression. Each of these dimensions has five associated severity levels and a weighted scoring procedure to create an EQ-5D-5L index score .
VAS has the appearance of a thermometer (vertical line with about 20 cm), whose scale ranges from 0 (worst imaginable health status) to 100 (best imaginable health status), with the individual having directly to mark what value s/he attributes to her/his current health status.
In this study, we have used the Portuguese version approved by the EuroQoL Group .
- To measure the satisfaction with life we asked the patient whether s/he agreed with the sentence “I am happy with my life”, in a 5-point scale from ‘completely disagree’ to ‘completely agree’. The social support question refers to the expected number of individuals the patient would be able to ask for help in case of need.
- The Hospital Anxiety and Depression Scale (HADS) is composed by 14 items, grouped in two subscales separately scored and measuring, respectively, anxiety and depression. Each item is answered in a 4-point Likert scale from 0 to 3 and it takes approximately 2-5 minutes to be completed . For each subscale, a score between 0 and 7 is considered normal, between 8 and 10 mild, between 11 and 14 moderate and between 15 and 21 severe.
In this study, we have used the Portuguese validated version .
- The sociodemographic information included gender, age, marital status, years of education and employment status.
We also asked physicians to provide clinical information about patient’s health status. This included age of migraine diagnosis, time of follow-up, headache frequency and characteristics, use of prophylactic treatment and types of drugs used in the past to prevent headache, as well as acute headache treatment. At last, we also collected information on patients’ comorbidities.
Statistically, one first decision we had to make was related to the use of non-parametric statistics, depending on whether mMIDAS-P index follows a normal distribution.
Therefore, we firstly raised the following hypothesis:
- H1: The distribution of mMIDAS-P index is normal
Reliability was assessed through intertemporal stability and internal consistency. To test the intertemporal stability we used the intraclass correlation coefficient (ICC). To avoid recall bias, eligible patients filled mMIDAS-P in two consecutive moments 3 weeks apart, the same interval used by the authors in their original validation study [4,5], as well as in some other countries. As defended by Koo, an ICC score lower than 0.50 corresponds to a weak correlation, between 0.50 and 0.75 to a moderate correlation, between 0.75 to 0.90 to a good correlation and a score higher than 0.90 corresponds to an excellent correlation . We also looked at the correlation factor between each item of mMIDAS-P and the whole scale.
The internal consistency was measured by the Cronbach’s alpha coefficient, where recommended values should be between 0.70 and 0.90 . For this study, two hypotheses were raised:
- H2: mMIDAS-P shows good intertemporal stability.
- H3: mMIDAS-P shows good internal consistency.
We followed the internationally defined methodology to validate QoL measures. Under this framework, we tested content, construct and criterion validity . As a translated Portuguese version already existed, we started by asking for a clinical revision performed by two neurologists, and a cognitive debriefing with 10 patients to guarantee the content validity, i.e., the relevance of the mMIDAS-P items. We formulated the following hypothesis:
- H4: mMIDAS-P is well accepted by clinicians and patients, not showing ambiguities, redundancies or lack of contents.
- Construct validity, addressing the theoretical concepts behind the measurement instrument, encompassed the structural validity and hypothesis testing . Structural validity was tested by exploratory factor analysis based on principal components estimates, having previously applied the sampling adequacy through the measure of sample adequacy Kaiser-Meyer-Olkin (KMO) and by Bartlett’s test of sphericity. A KMO smaller than 0.50, between 0.50 and 0.60 or between 0.60 and 0.70 is considered, respectively, unacceptable, poor or fair. Scores between 0.70 and 0.80, between 0.80 and 0.90 or higher than 0.90 are considered, respectively, average, good or very good . Bartlett sphericity test should have an associated significance of <0.001. For the selection of the number of factors, we followed the Kaizer criterion for eigenvalues greater than 1.
- For the hypothesis testing, we formulated several hypotheses with known-groups or subsamples based on sociodemographic and clinical variables. Should mMIDAS-P index be normally distributed, Student’s t-test was to be used for two independent variables and ANOVA for more than two independent variables. Should mMIDAS-P index be not normally distributed, the nonparametric Kolmogorov-Smirnov and Kruskal-Wallis tests were to be used. The following hypotheses were then defined:
- H5: The structure of mMIDAS-P maintains the unidimensionality.
- H6: mMIDAS-P scores are dependent from sociodemographic variables.
- H7: mMIDAS-P scores are dependent from clinical variables.
Criterion validity was assessed by comparing mMIDAS-P index with the ones obtained by HADS and EQ-5D-5L, considered as gold standard references. We also compared mMIDAS-P index with the results from the two additional questions of mMIDAS questionnaire. The statistical rule used was mainly based on Spearman’s correlation coefficient, but Pearson’s correlation was also computed to detect the role of possible outliers. As defined by Cohen, correlations smaller than 0.30 are considered weak, between 0.30 and 0.50 are moderate, and higher than 0.50 are considered strong .
By comparing mMIDAS-P with HADS and EQ-5D-5L, we expected to detect similarities and differences between dimensions, such as depression  and migraine pain . By comparing with the sixth and the seventh questions, we expected a positive correlation with headache frequency, measured by the number of days with headache, and with the headache intensity, measured by the average intensity of these headaches. Therefore, we tested the following hypotheses:
- H8: mMIDAS-P index correlates positively with the number of days with headache.
- H9: mMIDAS-P index correlates positively with the average intensity of headache.
- H10: mMIDAS-P index correlates with HADS dimensions anxiety and depression.
- H11: mMIDAS-P index correlates with both EQ-5D-5L index and VAS.