Abstract
We conducted a population-based headache questionnaire survey including questions on physician consultation for headache in Taipei, Taiwan from August 1997 to June 1998. The participants comprised 3377 subjects aged ≥ 15 years, of whom 328 (9.7%) had a diagnosis of migraine and 1754 (52%) had a diagnosis of non-migraine headache. Migraineurs had a higher physician consultation rate (once or more in the past year) than the subjects with non-migraine headache (54% vs. 31%, P < 0.0001). When frequency ≥ 10 times was taken as 10 times, the analysis showed that migraineurs consulted physicians more often than non-migraine headache subjects (2.36 vs. 0.96, P = 0.04). A small proportion of the subjects with either migraine (12%) or non-migraine headache (6%) accounted for 50% of total consultations within their groups. In addition to old age, low education levels, living in a rural area, migrainous features (nausea and photophobia), and work day loss, predictors of physician consultations also included ‘having been troubled with headache’ (odds ratio (OR) = 1.7) and co-morbidity with hypertension (OR = 1.8) or heart disease (OR = 2.2). Low copayment and unrestricted access to medical care, as well as cultural factors played an important role in the high consultation rates in our headache subjects. Moreover, this study found self-perception of headache impact and co-morbid illnesses were important factors affecting the decision to consult physicians about headache.
Introduction
Although there have been many large-scale prevalence surveys of headache in different countries, there have been few systematic analyses of the characteristics of subjects seeking physician consultations for headache (1, 2). Previous studies found that 19–80% of migraineurs never came to the attention of a physician (1–6), even though a significant proportion of them report high levels of pain and disability (1, 2).
Medical assistance-seeking behaviour is affected by disease severity, predisposition features (such as age and sex) as well as the type of health care system (7). In 1995, the government of Taiwan introduced a universal health insurance programme that covers all citizens. This National Health Insurance Programme removed the barriers to health care for all citizens (8). The programme is more of a social welfare programme than an insurance system. There is no upper limit to the number of physician consultations that can be made by an individual participating in the programme. Cultural factors interact with economic factors to give a distinctive character to out-patient care in Taiwan: short visits, many prescriptions and frequent follow-up visits (8). It was reported that the average number of physician consultations per person in Taiwan was 14.3 in 1997 (9).
The purpose of the present study was to determine the frequency of physician consultations for headache in both migraine and non-migraine headache sufferers and the predictors of physician consultations for headache using a population-based headache survey in the Greater Taipei Area of Taiwan.
Subjects and methods
The Greater Taipei Area Headache Survey was a population-based questionnaire interview conducted by trained interviewers from August 1997 to June 1998. The target population was subjects aged ≥15 years in randomly selected households in the Greater Taipei Area.
Study site and subjects
According to data on the 1996 year-end population structure in the Greater Taipei Area reported by the Department of Statistics, Ministry of the Interior, 1997 (10), the number of people aged ≥15 years was 4.6 million. This survey adopted a two-stage stratified random sampling method. Each Ts'un (village) or Li (subdivision of the district), which are basic administrative units in the Greater Taipei Area, was designated as the primary sampling unit, and each household in the primary sampling unit was designated the secondary sampling unit. In order to obtain at least 1000 respondent households, 1400 households were sampled according to the population structure. In the first stage of sampling, 10% of the Ts'uns or Lis were drawn systematically. Then, the sampled Ts'uns or Lis were allocated with proportions. In the second stage of sampling, household samples were selected systematically from the sampled Ts'uns or Lis in proportion to the size of the Ts'un or Li. The average sampling rate in the second stage was approximately 0.75%; and combining the two sampling stages, the overall combined sampling rate was about 0.075% of all the households in the Greater Taipei Area.
Questionnaire design
The questionnaire was composed of three parts. The first part investigated demographic characteristics. The second part was a clinically validated headache questionnaire (30 questions), which was designed to comply with the operational criteria of the headache classification of the International Headache Society (IHS), 1988 (11). A diagnosis of migraine was made according to the classification criteria of migraine without aura proposed by the IHS in 1988, except that those attacks with a duration of between 2 h and 4 h were also included (5). Before the large-scale survey was performed, 160 headache sufferers identified at the initial stage of the field study were sampled for validation of the use of the questionnaire to diagnose migraine. Among these headache sufferers, 142 (89%) were successfully interviewed by neurologists in a telephone semistructured interview within 8 weeks of questionnaire administration. The neurologists independently arrived at a diagnosis using IHS migraine criteria, without prior reference to the questionnaire data. The validity of the migraine diagnosis by the questionnaire was appropriate, with a κ value of 0.66, sensitivity of 76% and specificity of 91%. The third part of the questionnaire included self-awareness of health status, physician consultation for headaches, sleep quality and a validated mental health questionnaire: Chinese Health Questionnaire (CHQ) (12). The CHQ has 30 items and is used to screen for psychiatric illness, which is correlated with a score ≥10 (12).
Physician consultation for headache in the past year
Subjects with headaches were asked if they had consulted a physician for headache in the past year. They were asked how many times they went to see physicians and the specialty of their physicians. The exact number of consultations was categorized as 0, 1, 2,…8, 9, ≥ 10 times/year. We defined two groups of headache subjects according to their status of physician consultation in the past year: the consultation group as those who sought physician consultation at least once, and the non-consultation group as those who did not consult physicians in the past year.
Statistical analysis
Physician consultation behaviour was analysed in two groups: migraineurs and subjects with non-migraine headache. Student's t-test and χ2 tests were used for comparison when appropriate.
The predictors of physician consultation (≥ 1 time vs. 0 time/year) for headache were analysed in three steps. First, we grouped the potential predictors into five sets of variables, including (i) demographic variables: sex, age, educational level, marital status, household income, and living in Taipei City (urban) or County (rural) (the City is more prosperous than the County); (ii) headache profile: attack frequency, duration of headache attack, moderate or severe intensity as defined by the IHS (inhibition or prohibition of daily activities), migrainous symptoms listed by the HIS (locations, pulsatile headache, exacerbation by routine physical activities, nausea, vomiting, photophobia, and phonophobia) and presence or absence of questionnaire-diagnosed IHS migraine; (iii) functional impairment: days of work lost in the past year (work was defined in a broad sense including work, school, and doing chores for homemakers) (0, 1–3 days, 4–6 days, and ≥7 days), impairment (mild, modest, moderate, severe) of working abilities, daily living, or social functions; (iv) co-morbidity: presence or absence of a physician diagnosis of hypertension, diabetes, heart disease or depression, presence or absence of CHQ ≥ 10 based on a suspicion of psychiatric illness, and sleep quality (excellent, good, fair, poor, poorest); and (v) subjective awareness: self-awareness of health status and ‘having been troubled with headache’. Second, we performed five separate forward logistic regression analyses to screen predictors from each set of variables by using physician consultation as a dependent variable, and each set of variables as independent variables. The significant variables in each model were used as candidate predictors for final modelling. Third, a further forward multiple logistic regression model was designed using all of the significant variables chosen from the second step, to select a final suitable model for the likelihood of physician consultation after controlling for age, sex, and educational levels. The adjusted prevalence odds ratio (OR) with 95% confidence intervals (CI) for each predictor was calculated after controlling for all the other predictors in the final model.
All calculated P values were two-tailed and statistical significance was defined as a P value < 0.05. The reason why we did not apply the method of Bonferroni adjustment for multiple comparisons was because these headache variables were interrelated. In addition, a final logistic regression analysis was used to reduce the confounding effect among the variables.
Results
Participants
A total of 1211 households (86.5%) responded. Of the 4434 eligible subjects in the respondent households, 3377 subjects (76%) (F/M 1804/1573) completed the questionnaire. Compared with the demographic data for the population structure, there was a significantly higher ratio of female (53.42% vs. 50.05%, χ2 = 15.4, d.f. = 1, P < 0.001, goodness of fit test) and younger-aged group subjects (5 years for each age group) (χ2 = 132, d.f. = 11, P < 0.001, goodness of fit test) who completed the questionnaire.
Prevalence of questionnaire-diagnosed migraine
Of the 3377 participants, 2082 (62%) (50% men and 72% women) reported that they had at least one headache in the previous year. Among these 2082 subjects, 328 subjects (256 women and 72 men) had migraine with the 1-year prevalence of 9.7% (328/3377); whereas the others were designated as subjects having ‘non-migraine headaches’ (n = 1754) (52%).
Physician consultation rate
In the year prior to the questionnaire, 54% of the subjects with migraine and 31% of the subjects with non-migraine headaches reported having visited physicians once or more for their headaches (P < 0.0001). The consultation rates did not differ significantly between sexes (migraine M/F 56.9%/53.7%, P = 0.6; non-migraine headache 33%/29%, P = 0.07).
Frequency of physician consultation
If the consultation number was ≥10, it was assigned a value of 10 in the analysis. The resulting average number of physician consultations in the subjects with headache in the past year was 1.23, with a significantly higher frequency in migraineurs (2.36) than in non-migraine headache subjects (0.96, P = 0.042). Among the total participants, the average number of physician consultations was 0.73 times. Women consulted physicians significantly more often than men in the non-migraine headache group (1.05 vs. 0.85, P = 0.04), but not in the migraine group (2.39 vs. 2.26, P = 0.7).
Specialties of consulting physicians
The percentage of overall consultations by physician specialty was family practitioners (migraine vs. non-migraine headaches 29% vs. 14%), internists (17% vs. 10%), ENT doctors (14% vs. 7%), neurologists (12% vs. 5%), gynaecologists (4.9% vs. 1.4%), ophthalmologists (4.6% vs. 1.4%), allergists (1.2% vs. 0.5%), and others (2.7% vs. 1.5%).
Distribution of the physician consultations
We calculated the total consultation number as an index of healthcare utilization, where ≥10 consultations was assigned a value of 10 in the analysis. Figure 1 shows the cumulative percentage of total consultations vs. cumulative percentage of subjects with migraine (Fig. 1a) and non-migraine headaches (Fig. 1b). In the migraine group, 12% of subjects accounted for 50% of the total number of consultations for headache, and the consultation frequencies of these subjects were all ≥7 times/year. In contrast, only 6% of the subjects with non-migraine headaches accounted for 50% of the total number of physician consultations and the consultation frequencies of these subjects were all ≥4 times/year.
Cumulative percentage of total number of physician consultations for headache in subjects with migraine (a) and non-migraine headache (b). The cumulative percentage was calculated by ordering subjects from highest (≥ 10 times) to the lowest (0 times) number of physician consultation in the past year. (Note: the digits indicate the number of consultations).
Predictors of physician consultations (≥ 1 vs. 0 time/year)
Results of the predictor screening
The significant variables of physician consultations for headache screened from each set of variables by forward logistic regression analyses were summarized as follows: (i) demographics: age, sex, education levels and living in Taipei City or County; (ii) headache profile: frequency, moderate to severe headache intensity defined by IHS, nausea, photophobia; (iii) functional impairment: workday loss; (iv) co-morbidity: self-reporting physician-diagnosed hypertension and heart disease, and CHQ ≥ 10; and (v) subjective awareness: awareness of health status and having been troubled with headache (data not shown).
Results of the final fitting model
The above chosen variables were then put altogether into a forward logistic regression to select a final fitting model of predictors of physician consultation. Before predictor selection, age, gender and education levels were controlled. The final model chosen by the computer program is shown in Table 1. The most relevant predictors were workday loss, followed by having been troubled with headache, physician-diagnosed hypertension, moderate to severe headache intensity defined by IHS, headache frequency, physician-diagnosed heart disease, living in Taipei City or County, nausea and photophobia.
Final model results of predictor analysis of physician consultations for headache obtained by forward logistic regression analysis
Adjusted odds ratio for each predictor was calculated after controlling for all the other variables in this final model.
CI, Confidence interval; IHS, International Headache Society, 1988.
P < 0.05.
Discussion
Physician consultation rate and frequency
In this study, 54% of subjects with migraine and 31% of subjects with non-migraine headaches had consulted physicians because of their headaches in the past year. The average frequency of physician consultations was significantly higher in the migraine group than the non-migraine headache group. Compared with the findings of previous studies (range 10.0–36%) (1, 4, 5), the consultation rate in the migraineurs of this study was much higher. This was probably due to a combination of factors, including the low cost of copayments, unrestricted access to medical care, as well as cultural factors concerning medical assistance-seeking behaviour (8) in Taiwan. Considering that the average number of physician consultations per person was 14.3 in 1997 (9), consultations for headache accounted for 5.1% of all consultations among the study participants; i.e. one consultation in 20 was related to headache problems.
Physician consultation for headache was not uniformly distributed. A small proportion of the subjects with either migraine (12%) or non-migraine headache (6%) accounted for 50% of all physician consultations for headache within their groups. A similar distribution of work loss among the migraineurs was found in a previous study in which the more disabled half of migraineurs accounted for > 80% of all work loss (13). Targeting those headache patients in greatest need of medical care is therefore important from a medico–economic perspective.
Predictors of physician consultation for headache
The predictors of physician consultation found in the present study were similar to those of previous studies and included old age, workday loss (disability), headache severity, frequency, and migrainous features (1, 2, 5, 14). However, unlike previous studies, we also found that low socio-economic status (SES), co-morbidity with other physical disorders, and self-awareness were significant predictors of physician consultation for headache.
Headache sufferers with lower SES, i.e. those with lower education levels and living in Taipei County, were significantly more likely to consult physicians in the present study. A previous study found that lower SES was a predictor of a physician diagnosis of migraine among migraineurs in the Netherlands, where a social insurance health care system with unrestricted access was also provided (14). In contrast, a study done in the USA, where the majority of citizens were under private insurance systems, showed that a diagnosis of migraine was more likely in people with high-income levels (6).
Subject reports of a physician diagnosis of hypertension or heart disease were predictors of physician consultation for headache. The higher consultation rate for headache in patients with these two diseases was not caused by their association with migraine because, in our study, the patients with hypertension or heart disease did not have a higher prevalence of migraine (hypertension/no hypertension: 17%/15%, P = 0.6; heart disease/no heart disease: 18%/15%, P = 0.5). Nevertheless, these findings support the existence of Berkson's bias, which stresses the importance of co-morbidity of the other illnesses in Hospital patients (15). A popular concept in Taiwanese culture is that headache is frequently caused by high blood pressure, which may become an additional motivation for seeking physician consultation. In fact, a recent study also found an association between hypertension and severe headache in out-patients in Italy (16). However, we are unable to explain the finding of heart disease as a predictor. A causal relationship between co-morbid illness and physician consultations could not be determined in the present study because of the chance that the physician diagnoses of other diseases would be higher in those who sought physician consultation than in those who did not.
Having been troubled with headaches was a predictor of physician consultation (the second chosen variable), which was independent of the headache features, severity and disability. This finding indicates that perception of the symptoms of headache was a determinant of the decision to seek medical consultation. In fact, individuals with mild intensity headache who have visited many other doctors are commonly seen in our practice.
Methodology issue
Several methodological limitations of this study should be considered in interpreting the results. The response rate in this study was 86.5% for sampled households and 76% for eligible individuals, which is appropriate for a field study. However, since this study was designed and described as a headache survey, it is likely that sample selection inadvertently included a greater number of participants with headaches than is representative of the general population. Interviewers attempted to recruit all the eligible subjects in the respondent households to decrease this bias. Misclassification of headache subtype may have occurred in a certain proportion of study subjects because the sensitivity of the questionnaire-diagnosed migraine was 76% and specificity, 91%. We did not attempt to verify whether subject reports of physician consultations specific for headache or headache-related causes were accurate. In fact, this recall bias was even more likely to have affected results in subjects with multiple disorders, such as co-occurrence of hypertension and headaches. In addition, the accuracy of subject reports of physician diagnoses of illnesses was also not verified in the present study. In order to decrease the effect of recall errors, the participants were not asked to give an exact number of physician consultations if consultation was ≥10 times in the present study. However, setting ≥10 consultations as equal to 10 in the analysis underestimated the average number of consultations. In addition, in the analysis of predictors of physician consultation, we analysed each individual in the same household independently, which might overlook the family factor in a member's medical assistance-seeking behaviour.
Footnotes
Acknowledgements
This study was supported by National Health Research Institute grants (DOH86-HR-633, DOH87-HR-633, and DOH88-HR-633).